![]() |
|||||||||||||||||
![]() |
|||||||||||||||||
![]() ![]() ![]() |
![]() |
|
|||||||||||||||
![]() |
|
Comorbid Survival Among Elderly Male Participants of the Canada Health Survey: Relevance to Prostate Cancer Screening and Treatment
Abstract This study assessed the survival of a nationally representative sample
of older Canadian men, taking into account common comorbid conditions.
Mortality follow-up between 1978 and 1989 was conducted for male participants
of the Canada Health Survey who were at least 60 years of age at baseline.
The proportional hazards model and life table methods were used to examine
survival by comorbidity status. Comorbid conditions examined included
history of stroke and/or heart disease, high blood pressure, chronic
bronchitis or emphysema, diabetes and smoking status, but excluded cancer
because of small numbers. For those subjects aged 80 and older, comorbidity
was not a significant predictor of survival. A large portion of men
between the ages of 60 and 79, even those with pre-existing comorbid
conditions, survived at least 10 years after interview. In a clinical
setting, more detailed information on comorbid conditions can be obtained
to better estimate long-term survival. Notwithstanding, our findings
may have implications for the administration of population-based health
interventions (e.g. the use of prostate-specific antigen [PSA] blood
tests for the early detection of prostate cancer). In particular, our
results suggest that there may be little benefit in restricting access
to PSA screening based on survival probability in men under age 80. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Introduction Comorbid conditions are significant determinants of survival for most chronic diseases. The importance of classifying subjects by comorbidity has been demonstrated in patients with many chronic conditions, including diabetes mellitus,1,2 end-stage renal disease3 and breast cancer.4 The identification and control of comorbid conditions is also essential in the conduct of clinical trials. By properly identifying disorders that threaten survival, patients with an increased risk of death from comorbid condition(s) can be randomized separately from patients with a lower risk.5 Short-term studies often exclude patients with comorbid conditions, which limits their generalizability.6,7 For patients diagnosed with cancer, clinicians must often decide between administering a treatment that may compromise the patients' quality of life versus treating a malignancy in a less aggressive fashion. This is particularly relevant among older cancer patients who have a reduced life expectancy. Given the increased prevalence of comorbid conditions among the elderly, a better understanding of the influence of these conditions on survival may help determine the most appropriate cancer therapy. Assessing comorbid conditions may be important before using available screening tools, such as prostate-specific antigen (PSA) for the early detection of prostate cancer. For example, the American Cancer Society guidelines recommend PSA screening for men with at least a 10-year life expectancy.8 It may be that neither early diagnosis nor early treatment is necessary for men harbouring moderately or well-differentiated prostate cancer with limited life expectancies.9 Since the overwhelming majority of prostate cancer is diagnosed in men aged 60 and over,10 many of these patients have significant co-existing medical conditions that may contribute more in determining longevity than does prostate cancer. At present, it is unusual for clinicians to employ any rigorous methodology to predict longevity. Adjusting for comorbidity is also important in case-control studies of screening efficacy.9 The objectives of this study were to assess the individual and combined influence of comorbid conditions on survival among males aged 60 and older and to develop a comorbidity score that could be rapidly and easily applied to broadly determine the likely survival of an elderly man, using the Canada Health Survey.11
The Canada Health Survey The Canada Health Survey (CHS) was conducted in 1978 and 1979 to aid in the planning of health care, health promotion and disease prevention by examining lifestyle, biomedical and environmental risks to future health. It was designed to represent the non-institutionalized Canadian population, with the exception of the approximately 3% of the population that resided in the territories, Indian reserves and remote areas as defined by the Canadian Labour Force Survey. Sampling was stratified by province, then by area (major city, urban and rural).11 A total of 12,218 dwellings were selected to participate in the CHS; of these, 86% took part in the interview component of the survey.11 A physical measures component was administered to roughly 25% of eligible households, collecting measurements of blood pressure, cardiorespiratory fitness, height and weight. Three forms were used in the interview section of the CHS. The first, the Household Record Card, recorded particular characteristics of the dwelling and residents. The second form, the Interviewer Administered Questionnaire (IAQ), collected data that, in general, required probing by an interviewer. For this purpose, proxy information was gathered from a suitable member of the household. The Lifestyle and Your Health Questionnaire (LHQ) gathered information that could be sensitive and only reliably provided by the person involved. Due to its content, this third questionnaire was administered only to persons 15 years of age and over. There was a 16% non-response rate for the LHQ, resulting from subjects who responded to the IAQ but who did not provide a valid response to at least one compulsory item on the LHQ. Follow-up of the Cohort We determined the mortality history of the cohort by linking the CHS file to the National Mortality Database maintained by Statistics Canada. Record linkage for deaths occurring between 1978 and 1989 was performed using an iterative probabilistic weighting scheme.12 Questionable links were resolved manually by inspecting death certificates. Previous studies have demonstrated the ability of these record linkage techniques to identify vital status in longitudinal studies.13,14 Selection of Subjects We examined the mortality experience only of males who were at least 60 years old at the time of interview. In total, 1939 subjects of the CHS met this inclusion requirement. Eighty-four individuals with a previous diagnosis of cancer (excluding non-melanoma skin cancer) were excluded from the analysis. We felt that assessment of the influence of cancer as a comorbid factor could not be made due to the small number of cases and the unavailability of information on the site of cancer diagnosis. The comorbid conditions examined included history of high blood pressure, heart disease, emphysema or bronchitis, stroke and diabetes. For each subject this information was obtained from both the IAQ and LHQ components of the survey. Each comorbid condition was coded using an indicator variable. Cigarette smoking information was obtained from the LHQ interview. Approximately 15% of the 1939 subjects who completed the IAQ were dropped from the analysis due to non-response on the LHQ. Subjects for whom smoking status was not available were also excluded from the analysis, resulting in a loss of 9.7% of the remaining subjects. All statistical analyses were performed on the remaining 1417 subjects. Statistical Analysis We used the proportional hazards model15 to assess the role of comorbidity on survival. This model assumes that those patients with comorbid ailments have an increased risk of mortality that remains constant over the follow-up period. In our analyses, the assumption of proportionality was formally tested by examining survival curves generated using the actuarial life table and by testing the significance of a time covariate in the regression model. Subjects were classified as current, former or never smokers. All other comorbid conditions were coded as a binary variable (0 = absent, 1 = present). Two series of risk estimates were produced with the proportional hazards model. Unadjusted relative risks assessed the risk of mortality for a patient with a given comorbid condition, independent of the presence of other comorbid factors. The second series of risk estimates was further adjusted for the influence of all other comorbid factors and smoking. Both series of risk estimates were adjusted for differences in age at interview by including a categorical variable, denoting five-year age groups, into the proportional hazards model. The adjusted relative risks estimated from the proportional hazards model were used to create a variable that summarized the overall influence of comorbidity on survival. This summary variable was then entered into a proportional hazards model with age to create an index of survival that combined both age and comorbidity. Survival curves were then plotted by the age-comorbidity score, using the actuarial life table method. Results The average age of the subjects included in the analysis was 68.6 years, and 80% of them were either current or former smokers (Table 1). The proportion of patients aged 60-69, 70-79 and 80 or older who reported at least one comorbid condition besides smoking was 48%, 58% and 68%, respectively. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
As anticipated, long-term survival rates decreased with increasing age (Table 2). The survival rates of those men aged 80 or older were particularly low, with only 26% surviving 10 years of follow-up. Poorer survival was also observed for those subjects reporting comorbid conditions. Among males aged 60-79, after adjustment for the presence of other comorbid factors, the risk of mortality for current smokers was 1.57 relative to those who had never smoked (Table 3). Similarly, the risks of mortality for subjects with a prior history of heart disease, chronic bronchitis or emphysema, diabetes, high blood pressure and history of stroke were 1.52, 1.57, 1.17, 1.28 and 1.53, respectively, when compared to those with no history of the corresponding comorbid factor (Table 3). | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Because comorbidity was not significantly predictive of risk among men aged 80 and older, we calculated age-comorbidity scores only for men aged 60-79. A weighted index was created based on the adjusted estimates of relative risk for men aged 60-79. For those conditions whose relative risk was less than 1.5, the subject was assigned a weight of one if the comorbid condition was present and zero, otherwise. Similarly, conditions with a relative risk greater than or equal to 1.5 were given a weight of two or zero. The assigned weights for each of the comorbid conditions are listed in Table 4. For each subject, the weights of all the comorbid conditions were summed, yielding a summary measure of comorbidity with a range between zero and eight. A proportional hazards regression analysis was subsequently performed using the summary variable for comorbidity and age as covariates. The risk associated with each five-year increase in age was approximately 1.8 times the risk associated with increasing the comorbidity index by one. An age-comorbidity index of survival was constructed, using the approach outlined by Charlson.4 Each five-year age group after age 60-64 was assumed to contribute two points of risk. The age-comorbidity score was the sum of the individual scores associated with age and comorbidity (Tables 4 and 5). For example, a 77-year-old cigarette-smoking subject would have a comorbidity score of two, an age score of six and an age-comorbidity score of eight. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Decreased survival was observed with increased levels of comorbidity for subjects aged 60-69 and 70-79 (Figures 1 and 2). There was no clear gradient associated with comorbidity for subjects aged 80 and older. Ten-year survival rates for men in this age group ranged from 14% to 35% across categories of comorbidity (Figure 3). The 10-year survival rate for these men with an age-comorbidity score of zero was 94% (Table 5). The relative risk of mortality increased dramatically with higher age-comorbidity scores. Those having an age-comorbidity score of at least nine were approximately 15 times more likely to die during follow-up (Table 5). Kaplan-Meier survival curves are provided in Figure 4 for categories of age-comorbidity scores. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Discussion The index derived in this study permits the classification of subjects into risk categories that take into account common comorbid conditions, such as smoking and age. Thus, for an elderly man who develops a disease with a poor prognosis, the most appropriate therapy to administer might be one whose primary goal is to maximize quality of life and for which the reduction of long-term consequences of the disease is of secondary interest. Conversely, the proposed index would allow the identification of patients for whom aggressive therapy might be more suitable. Although the model performed quite well in predicting survival, its utility in predicting survival will need to be validated using other population-based cohorts. Another use of comorbidity scores would be to assess the merits of treating or screening for particular health problems. For example, it is recommended that men with a life expectancy of less than 10 years be advised that both screening and treatment for prostate cancer are unlikely to be beneficial and may decrease quality of life.8,17 An increasing number of physicians are ordering the PSA screening test for men with neither signs nor symptoms of prostate cancer.17 The data used in this study offer several advantages compared with prior research on the role of comorbidity on survival. Previous studies have relied on hospital or clinic data. Because the CHS was designed to represent the Canadian population, the present study should have avoided any potential bias that could have arisen because of referral patterns. The large sample size of the CHS allowed for the analyses to be limited to males 60 years of age and older. The influence of comorbidity on survival in this age group is of particular interest right now, due to the much higher prevalence of chronic diseases in this population coupled with current controversies surrounding the efficacy of PSA testing. Although the Canada Health Survey was designed to draw from a representative sample of Canadians,11 a sizeable number of subjects were dropped from our analyses due to missing data. The 10-year survival rate for men excluded from analyses was lower than for those subjects who were retained (55% vs 63%). The difference in survival between these two groups can be partly attributed to differences in age. The mean age of those men removed from analyses was approximately 1½ years greater than the mean age of those included. The net effect of excluding these subjects from our analyses is that our survival estimates may slightly overestimate the mortality experience of the older male Canadian population. The prognostic value of common comorbid conditions among males aged 60-79 is presented in this study. Among those men aged 80 or older, comorbid conditions were not significant determinants of survival. A more refined characterization of survival could have been done had other baseline characteristics been included in the regression analyses. For example, other important predictors of survival for people with diabetes include type of diabetes (i.e. insulin- or non-insulin-dependent) and onset of diabetic nephropathy.18 Similarly, the risk of mortality for smokers is associated with duration of use, daily consumption and depth of inhalation.19 However, we have shown that the scores derived even from simple, unvalidated self-reports on comorbidity status were significantly predictive of survival. It is likely that the presence of comorbid conditions were underreported by participants of the CHS. As a result, the survival rates presented probably overestimate the true survival of males with comorbid conditions and underestimate survival for males with no comorbid conditions. The smoking behaviour of some subjects likely changed during the follow-up period. However, the resulting effect on risk of mortality by smoking status should not be substantial. The vast majority of smokers initiate smoking during their teenage years; older smokers usually have been addicted for a longer period and, consequently, are less likely to change consumption patterns.19 Additional comorbid conditions other than smoking also may have developed during the follow-up period. Such changes do not detract from the comorbidity and age-comorbidity indices, since their value lies in the ability to predict survival given only baseline measures. Since the initiation of follow-up of the CHS, many improvements have been made in the treatment of chronic diseases and comorbid factors. For example, new hypertensive agents and refinements in therapy have reduced long-term stroke and cardiac complications.20,21 With continuous progress in the control and prevention of chronic disease, it follows naturally that the long-term survival rates of males who are now over the age of 60 should be higher than the rates of those on whom the analysis was based. Notwithstanding a lack of endorsement from the Canadian Urological Association,22 PSA screening is widespread in Canada.23 Prostate cancer is controversial, in part because some prostate cancers detected by screening will be indolent and, especially among older men with comorbid conditions, will not affect survival. However, our results demonstrate that a large number of men between the ages of 60 and 79, even those with pre-existing comorbid disease, are likely to survive at least 10 years. Consequently, efforts to restrict access to PSA screening based on survival probability are unlikely to substantially affect the overall costs of mass screening.
Author References |
![]() |
|||
![]() |
Last Updated: 2002-10-20 | ![]() |