Colorectal carcinoma is the fourth most common nonskin malignancy in the U.S., with approximately 130,700 new cases expected in 2000, but it is the second leading cause of cancer deaths after lung carcinoma, with 56,300 deaths expected in 2000.1 As advances in treatment are realized, the number of colorectal carcinoma survivors may increase over time. Currently, there are approximately 1.24 million survivors of colorectal carcinoma living in the U.S.2 The annual direct medical care costs for colorectal carcinoma has been estimated at $6.5 billion (in 1990 dollars), second only to breast carcinoma.3
Despite the prevalence and burden of colorectal carcinoma to patients, their families, and society, to our knowledge there is relatively little information regarding health-related quality of life (HRQOL) among individuals with colorectal carcinoma in representative community settings. This is particularly true for individuals who have survived beyond the initial treatment period. HRQOL information on colorectal carcinoma survivors is important if one wishes to evaluate the impact of the disease and its treatments beyond survival, and to understand how HRQOL is influenced by age and stage at diagnosis. In addition, HRQOL evaluations may detect chronic problems that linger or occur de novo months to years after the therapy has ended.4
In this article, we report the results of a survey of HRQOL for a representative community-based sample of individuals with colorectal carcinoma as a function of their stage at diagnosis and time since diagnosis. The survey was part of a larger effort to estimate the cost-effectiveness of screening for colorectal carcinoma using alternative screening modalities.
MATERIALS AND METHODS
- Top of page
- MATERIALS AND METHODS
The overall objectives of the study were to evaluate both general health status and several disease specific dimensions of illness in people with colorectal carcinoma, focusing on how these measures of HRQOL compared for colorectal carcinoma survivors with different TNM stages at diagnosis and times since diagnosis. Our target population was individuals who had survived at least 1 year from diagnosis. To accomplish these objectives, we surveyed 173 colon carcinoma survivors identified through the Cancer Surveillance System (CSS), a registry of cancer cases for residents of western Washington State. The CSS is part of the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program. The SEER registry tracks cancer cases from 11 regional reporting areas throughout the U.S.5, 6
When considering a lifetime history of an individual with colon carcinoma for cost-effectiveness analyses, it is useful to construct a hypothetical stream of costs and outcomes (quality of life, survival) over time. Previous models of the lifetime costs of cancer divide costs into time periods that include prediagnosis, initial care, continuation, and terminal care.7–9 Models of HRQOL for chronic illnesses also have been similarly constructed.10 Although longitudinal sampling is desirable for tracking health outcomes over time, in chronic illnesses such as colon carcinoma there are several practical limitations to this approach. Individuals may be reluctant or unable to fill out survey forms during the initial and terminal phases of their disease. Tracking can be difficult for long term illnesses when subjects move out of the area. Accordingly, our strategy used cross-sectional sampling that focused on individuals in the “continuing care” phase of their illness. We define continuing care as > 12 months from the time of diagnosis. Our hypothesis was that the HRQOL would be higher for individuals who were diagnosed at an earlier stage of colorectal carcinoma, particularly during the continuation phase of their illness, and for those with a longer time since diagnosis. In addition, to contrast HRQOL for those individuals in a continuing care phase of illness with those in a terminal phase of illness, we retrospectively examined the database and identified those individuals who died within 1 year of being interviewed. Finally, we sought to identify the health dimensions that are most influenced by colon carcinoma, specifically in the continuation phase of illness, to determine whether those dimensions were affected by stage at diagnosis and whether they changed over time.
Subjects were selected from the CSS registry if colon or rectal carcinoma was their first recorded malignancy and if they were older than age 19 years at the time of diagnosis. Subjects were sampled from the registry to represent all colorectal carcinoma TNM stages at diagnosis except Stage 0 and 4 time periods from the date of initial diagnosis: 1) 13–24 months; 2) 25–36 months; 3) 37–60 months; and 4) > 60 months. Sampling was conducted using an algorithm that was designed to maximize the accuracy of the estimates of lifetime HRQOL, by colorectal carcinoma TNM stage at diagnosis and time since diagnosis, using cross-sectional data.11 The sampling algorithm entailed a two-phase procedure. During the first phase, survivors diagnosed with TNM Stage I–IV colorectal carcinoma were selected from the database. At the end of Phase 1, HRQOL was analyzed after stratifying respondents by stage at diagnosis. The number of subjects selected in each TNM stage for the second phase of the study was based on the degree of variation in HRQOL within stages found during the first phase. More subjects were sampled from stages with higher degrees of variation in HRQOL. Human Subjects Committee approval was obtained for all activities.
The surveys were conducted in two phases, the first between April and November of 1996 and the second between May and December of 1997. In both phases, permission was obtained from the subject's physician prior to contacting the individual. Subjects initially were contacted via a recruitment letter containing a set of questions that screened for eligibility to participate in the survey. Nonresponders were contacted and screened for eligibility by telephone. Subjects were considered ineligible if they (or their physician) denied that they had been diagnosed with colorectal carcinoma; if they had problems with hearing, speaking, or understanding English; or if they had a second, noncolorectal carcinoma-related malignancy. If the physician denied permission to contact the subject or the subject refused to participate during the screening process, that subject was replaced with another individual from CSS who most closely matched his or her stage at diagnosis and time since diagnosis. Resampling was necessary because low response rates were anticipated for those subjects with advanced disease and those who were several years from the date of diagnosis (primarily due to moving out of the region).
During the first phase of the study, colorectal carcinoma survivors were asked to come to an interview site in Seattle to complete a series of self- and interviewer-administered surveys. At the interview, trained interviewers asked individuals to comment on the clarity and acceptability of the questions in the self-administered survey. Questions that were difficult to understand or objectionable to a number of interviewees were dropped from the self-administered survey used in the second phase. The revised self-administered survey was mailed to colorectal carcinoma survivors identified from the CSS for the second phase of the study. Nonresponders to the mailed survey were contacted by telephone. Those who could not be reached or refused to participate were replaced with an individual from the database who most closely matched the nonrespondent's stage at diagnosis and time since diagnosis.
We report here results from the questions that were identical in both self-administered surveys.
The surveys included a health question for global rating of health (poor, fair, good, very good, and excellent), a disease specific measure of colorectal carcinoma-related quality of life, and a generic measure of HRQOL that records individual health state preferences, also known as utilities. Utility values range from 0–1 with the anchor points reflecting death and optimal health. Utilities are recommended for adjusting survival to derive quality-adjusted life-year (QALY) estimates for cost-effectiveness analyses.12 Two validated survey instruments were included in the questionnaire: the Functional Assessment of Cancer Therapy Scales for colorectal carcinoma (FACT-C) and the Health Utilities Index (HUI) Mark III. The FACT-C is a disease specific HRQOL instrument that measures the health status of patients with colorectal carcinoma. Dimensions of HRQOL measured by the FACT-C include physical, emotional, functional, and social well-being and satisfaction with the physician-patient relation.13 Questions cover general cancer-related issues and colorectal carcinoma specific issues. Scores for individual dimensions are added to produce a summary score ranging from 0 (worst possible) to 112 (best possible). The HUI is a generic multiattribute, preference-based system for assessing HRQOL in a wide variety of chronic health states of varying severity. Questions cover eight health-related dimensions (vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain/discomfort). Individual dimensions are weighted according to values derived from the general population and summed to produce utility values ranging from 0–1.14, 15
The survey instrument also included several demographic questions, a question regarding whether the respondent wore a colostomy appliance, and questions of whether the patient received surgery, chemotherapy, or radiation therapy for their colorectal carcinoma in the past month. Respondents also were asked to list their chronic illnesses other than colorectal carcinoma.
Respondents from the two survey groups first were analyzed to determine whether the populations differed significantly in demographic or cancer-related characteristics. Multiple pairwise comparisons were performed to determine the degree of similarity of the populations sampled from CSS for Phase 1 and Phase 2. The combined sample then was analyzed using logistic regression to determine whether certain characteristics predicted nonresponse to the invitation to complete the survey.
After performing the tests of homogeneity, subjects from the combined database were stratified by TNM stage at diagnosis. HUI and FACT-C summary scores were plotted and trends in HRQOL over time were determined for each stage and for all stages by fitting curves to the plots using a smoothing technique. The smoothing technique fits curves to data points locally so that at any point on the curve the value depends only on the observations at that point and at neighboring points. Smoothed curves were plotted using the “supsmu” command in S-PLUS.16
Because the HRQOL data were highly skewed, nonparametric tests were used to evaluate the unadjusted, stratified data. Differences in HRQOL across colorectal carcinoma TNM stages were compared using the Kruskal–Wallis test.17 Differences in HRQOL after stratifying based on whether the subject wore a colostomy appliance were tested using the Mann–Whitney U test. Average stage specific scores for the HRQOL dimensions of both survey instruments also were compared using the Kruskal–Wallis test.
Multivariate regression was used to determine the adjusted effects of stage at diagnosis and time since diagnosis on global HRQOL and disease-related HRQOL for those in the continuation phase of illness. The dependent variables for each of the regression models were the HUI summary scores (possible range, 0–1) and FACT-G summary scores (possible range, 0–112), respectively, and subscale summary scores for each survey measure. Independent variables included age, gender, race (white or nonwhite), low income status (family income < $20,000/year), marital status, time since diagnosis (months), and TNM stage at diagnosis (indicator variables representing each stage). Low education status was not included in the regression due to its high correlation with low income status. Interaction terms for stage at diagnosis and time since diagnosis were added to allow change in HRQOL over time to differ from stage to stage.
- Top of page
- MATERIALS AND METHODS
The purpose of this analysis was to determine how HRQOL varies as a function of stage at diagnosis and time since diagnosis for colorectal carcinoma survivors. Our graphic results indicate that for all stages, the impact of colorectal carcinoma on HRQOL is greatest in the first 2–3 years after diagnosis. With the exception of those who are diagnosed with Stage IV disease, those who survive > 3 years appear to experience a relatively high general HRQOL.
The subscale scores for both survey instruments indicated that the emotional component of HRQOL improves significantly over time, although this dimension was not the factor with the greatest decrement in this patient group. The deficits in hearing and vision, measured by the HUI but not by the FACT-C, are likely more a function of age-related rather than disease-related factors. Functional and social well-being, measured by the FACT-C but not the HUI, were impacted most for this group, and appeared to change little (i.e., improve) as time from diagnosis lengthened. This may indicate that colorectal carcinoma survivors suffer long-lasting reductions in these dimensions. Results on the pain dimension of the HUI suggest that pain may be a substantial problem that does not improve with time for colorectal carcinoma survivors. It is important to note that a pain question was included in the FACT-C; however, it was one component of the physical well-being dimension, which was not as impacted as others for this patient group. Finally, we noted that low income status was associated with poorer outcomes on the emotional dimensions of both surveys, and with the pain dimension on the HUI and physical dimension of the FACT-C (which includes pain). Because colorectal carcinoma-related pain is unlikely to be different across socioeconomic strata, this consistency may indicate that low income persons with colorectal carcinoma may have inadequate access to health care services related to pain management.
Although these trends are apparent from the plotted data, we did not find HRQOL to be significantly different as a function of stage at diagnosis or time since diagnosis in categoric or multivariate regression analysis. We believe this discrepancy is in part explained by the nature of how HRQOL may change over time for patients with colorectal carcinoma, and is not accounted for by the linearity of the time term in the multivariate regression. Piecewise analysis of HRQOL over short time periods would reduce the confounding effect of the time-treatment interaction. The number of subjects in this relatively small study was not sufficient for such analysis. In addition, the lack of apparent difference also is likely to be a function of our small sample size and the limitations that are inherent in cross-sectional sampling designs.
HRQOL for patients with colorectal carcinoma is likely affected both by the burden of the disease itself and the treatment regimens that are applied for the disease. Conceptual models of illness suggest that HRQOL often falls during the treatment phase, then recovers over weeks to months after therapy is completed. This study focused on individuals hypothesized to be in “continuing care” (beyond their first year since diagnosis) and, to a lesser extent, “terminal” (last year of life) phases of illness. We found that HRQOL remained highly variable and changing several years into what is typically considered a “continuing care” phase. Individuals in their last year of life appeared to experience marked declines in HRQOL compared with those who survived beyond this time period. Whether the decline in HRQOL and death is due to progressive colorectal carcinoma or other factors cannot be determined from this survey.
It is possible that much of the variation in HRQOL during the first several years after diagnosis can be accounted for by individuals who were experiencing disease recurrence. Of the 16 subjects in the sample who noted that they recently had received colorectal carcinoma-specific therapy, only the postsurgery group had a significantly different HRQOL compared with those who did not report recent therapy. However, active treatment may be an insufficient proxy for detecting those who are experiencing disease recurrence. We recommend considering other approaches to differentiate those who are in disease remission from those with recurrent disease when measuring HRQOL in cancer survivors.
Cross-sectional sampling designs are subject to response biases. We anticipated that more frequent resampling would be necessary for those subjects with advanced stage disease and a distant history of colon carcinoma. However, comparison of respondents and nonrespondents indicated that advanced age was the only factor significantly associated with nonresponse. In addition, the retrospective analysis did not demonstrate that nonrespondents were significantly more likely to be in their last year of life. Nevertheless, we caution that the possibility still exists that there was a response bias favoring healthier subjects. The robustness of these results can be evaluated only with additional surveys of colorectal carcinoma survivors. Notably, generic HRQOL surveys of breast and prostate carcinoma survivors also have shown that their overall quality of life is similar or higher than their age-matched peers without cancer.19–21 This could be interpreted as a widespread selection bias problem, or an indication that cancer survivors in general find that their experience has in some ways enhanced their appreciation of day-to-day life.
A second sampling issue is that the population was largely white, elderly, middle class, and married. Therefore, caution is advised before extrapolating these results to groups with different sociodemographic characteristics.
Although population norms derived directly from the HUI are not available, Erickson et al. constructed a Health Utility Index score from questions asked within the NHEFS (NHEFS-HUI).18 Average HRQOL calculated from the NHEFS-HUI ranged from an average of 0.81 for persons ages 55–60 years to 0.70 for persons ages 75–80 years, compared with an average of 0.84 for the respondents to the current survey.
Other investigators have evaluated HRQOL in colorectal carcinoma survivors. Using the time tradeoff technique, Smith et al. obtained utility weights of 0.8–1.0 for Dukes Stage C patients who had just undergone surgery for their colorectal carcinoma but had not yet begun chemotherapy.22 Using a cancer specific HRQOL instrument in a cross-sectional study, Schag et al. found that the HRQOL improved for colon carcinoma survivors as their time from diagnosis lengthened, in contrast to survivors of lung carcinoma and prostate carcinoma, in whom HRQOL remained stable or declined, respectively.4
Our data suggest that colostomy did not affect overall HRQOL significantly. In a review article, Sprangers et al. suggested that survivors with colostomy appliances report substantially more problems with physical, psychologic, social, and sexual functioning than colorectal carcinoma survivors with preserved anal sphincter function.23 They note that body image issues may be particularly important for female patients and younger patients. There were relatively few young subjects in our sample.