Comorbidity and age as predictors of risk for early mortality of male and female colon carcinoma patients

A population-based study


  • Rosemary Yancik Ph.D.,

    Corresponding author
    1. Geriatrics Program, National Institute on Aging, National Institutes of Health, Bethesda, Maryland
    • National Institute on Aging, National Institutes of Health, Gateway Building, Suite 3E327, 7201 Wisconsin Avenue MSC 9205, Bethesda, MD 20892-9205
    Search for more papers by this author
  • Margaret N. Wesley Ph.D.,

    1. Information Management Services, Inc., Silver Spring, Maryland
    Search for more papers by this author
  • Lynn A. G. Ries M.S.,

    1. Cancer Control Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Richard J. Havlik M.D., M.P.H.,

    1. Epidemiology, Biometry, and Demography Program, National Institute on Aging, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Sherrill Long B.A.,

    1. Information Management Services, Inc., Silver Spring, Maryland
    Search for more papers by this author
  • Brenda K. Edwards Ph.D.,

    1. Cancer Control Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Jerome W. Yates M.D., M.P.H.

    1. Roswell Park Cancer Institute, Buffalo, New York
    Search for more papers by this author



Colon carcinoma primarily affects older-aged persons 65 years and older. Seventy-five percent of the incident tumors affect persons in this age group. Because of their advanced age, older patients already may be coping with other concomitant major physical illnesses. This article documents preexisting diseases in older colon carcinoma patients at diagnosis and evaluates the effects of their comorbidity burden on early mortality.


Prevalence of comorbid conditions was assessed by a retrospective medical records review of an age-stratified random sample of male and female patients aged 55-64 years, 65-74 years, and 75+ years (males, n = 799; females, n = 811). Data were collected on comorbidity by the National Institute on Aging (NIA) and National Cancer Institute (NCI) and merged with NCI Surveillance, Epidemiology, and End Results (SEER) tumor registry data.


Hypertension, high impact heart conditions, gastrointestinal problems, arthritis, and chronic obstructive pulmonary disease emerged as the most prominent comorbid conditions in the NIA/NCI SEER study sample. The prevalence of comorbidity in the number and type of conditions similar for both men and women (e.g., 40% of each gender had ≥ 5 comorbidities). Within 2 years of diagnosis, 28% (n = 454) of the patients had died. The number of comorbid conditions was significant in predicting early mortality in a model including age, gender, and disease stage (P = 0.0007). Certain comorbidities, classified as "current problem," added significantly to a basic model (e.g., heart problems, alcohol abuse, liver disease, and deep vein thrombosis).


Although disease stage at time of diagnosis of colon carcinoma is a crucial determinant of patient outcome, comorbidity increases the complexity of cancer management and affects survival duration. Cancer control and treatment research questions should address comorbidity issues pertinent to the age group primarily afflicted with colon carcinoma (i.e., the elderly). Cancer 1998;82:2123-2134. © 1998 American Cancer Society.

Colon carcinoma is among the three leading sites of cancer incidence and mortality in the United States after lung and prostate carcinoma for men, and lung and breast carcinoma for women.1, 2 In 1997, it is estimated that 94,100 persons will have been diagnosed with colon carcinoma; deaths caused by this tumor are estimated to reach as high as 46,600.2

Colon carcinoma is especially common for men and women aged ≥ 65 years. Seventy-five percent of the incident tumors affect persons in this age group. The National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) data show that the overall age-adjusted cancer incidence rate for persons in this age group is 20 times greater than the rate for persons younger than 65 years (235.8 per 100,000 population compared with 12.0 per 100,000 population).1 Thus, colon carcinoma affects a large number of individuals who, because of advancing age, already may be coping with concomitant physical illnesses such as heart disease, chronic obstructive pulmonary disease (COPD), arthritis, diabetes, hypertension, and other major health problems. Comorbidity (i.e., preexisting diseases) may provoke adverse consequences and affect conventional approaches in management of the tumor. The patient may not be able to undergo optimal treatment intervention for colon carcinoma (i.e., surgery).

We know little regarding the extent of the comorbid disease burden of colon carcinoma patients or how it affects the diagnosis and outcome of this malignancy. Whether the concurrent diseases are active symptomatically or controlled (e.g., by drugs, changes in life-style, and/or surgery), they become a part of the physical makeup of an aged individual and are likely to remain so for life.3 A malignancy superimposed on extant age-related conditions in older patients generates a clinical dilemma for treatment of the cancer. Older cancer patients also may have normal age-related changes and residual disease effects that interact with tumor progression.4

The National Institute on Aging (NIA) Epidemiology, Demography, and Biometry Program and the NCI SEER Program conducted a collaborative investigation to ascertain the comorbidity burden and its impact on older-aged patients. Colon carcinoma was one of seven tumors chosen for the NIA/NCI SEER Study, which began in 1992. In this article, we document the comorbidity in older colon carcinoma patients and evaluate the influence of various diseases on cancer survival. We focus on the role of comorbidity in early mortality. Guiding research questions were: 1) What is the comorbidity burden of older colon carcinoma patients? 2) Does the comorbidity burden of colon carcinoma patients correlate with their disease stage? and 3) Does the comorbidity burden contribute to early mortality of the patient?


Data sources are the NCI SEER tumor registry data on an age-stratified random sample of male and female colon carcinoma patients merged with information obtained through a thorough retrospective review of their medical records to assess comorbid conditions. Colon carcinoma patients from six SEER registries (Iowa; New Mexico; Utah; San Francisco-Oakland, California; Atlanta, Georgia; and Seattle, Washington), diagnosed with colon carcinoma between January 1 and December 31, 1992, were randomly sampled in three age strata (55-64 years, 65-74 years, and 75 years and older) by gender. Patients in the youngest age group served as a comparison control with the two older age groups. Patient selection and data collection methods of the NIA/NCI SEER Study on Comorbidity and Cancer in the Elderly have been published elsewhere.5, 6

Trained abstractors, following a specially designed abstract format, study guidelines, and established definitions, reviewed the entire medical record for detailed information on comorbid conditions, signs and symptoms of cancer, and admission medications. Major sources for the comorbidity data in the medical record were physician notes, anesthesia notes, nursing notes, discharge summaries, and laboratory reports. This information was matched with routinely obtained SEER tumor registry data (age, gender, extent of disease at diagnosis, treatment, and survival). This article analyzes data on 1610 patients with colon carcinoma (799 males and 811 females).

The primary data collection instrument listed 27 major comorbid conditions and included space for recording additional comorbidities that were later coded. The array of comorbidities and subcategories are listed in Table 1. Classifications recorded on the abstract form are grouped according to number of conditions present, level of current or historic impact of the condition, and no information available. "Impact" codes were obtained by NIA/NCI SEER Study abstractors on whether diseases were: 1) a current medical management or diagnostic problem (designated as "‡" in Fig. 1 (7K)) and 2) noted in the medical record but not a current management or diagnostic problem (designated as "§" in Fig. 1 (7K)). If no information on these diseases was present in the medical record, it was assumed the patient did not have this particular problem. Key variables-Total Comorbidity and Life Threat-were derived from these data.

Figure 1.

Comorbid conditions according to presumed Life Threat Risk High to Negligible. ‡: under acrtive management; §: no current management/history only; TIA: transient ischemic attack; Hx: history.

Table 1. List of Comorbid Conditions in the Colon Carcinoma Study Sample (NIA/NCI Collaborative Study)
  1. NIA: National Institute on Aging, NCI: National Cancer Institute, GI: gastrointestinal, SEER: Surveillance, Epidemiology, and End Results.

  2. Comorbidities assessed in the National Institute on Aging/National Cancer Institute Surveillance, Epidemiology, and End Results Collaborative Study on Comorbidity and Early Diagnosis of Cancer in the Elderly, 1997.

Alcohol Abuse
Alzheimer's Disease
       joint disease
  Other arthritis
  Inflammatory arthritis
  Rheumatoid arthritis
  Polymyalgia rheumatics
Chronic obstructive pulmonary
  Chronic bronchitis
Deep vein thrombosis
  Currently receiving insulin
  Current problem without insulin
Eye problems/ophthalmic disease
Gallbladder problems
Gastrointestinal problems
  GI hemorrhage
Heart-related conditions
  Cardiac arrest
  Cardiovascular disease
  Congestive heart failure
  Myocardial infarction
  Valve disease
  Other heart problems
    Pulmonary embolism
    Unspecified peripheral
       vascular disease
    Giant cell arteritis
    Structure of artery
    Unspecified heart problems
Lipid problems
  Other hyperlipidemia
Liver disease
Mental health problems
  Paranoid residual unspecified
  Bipolar affective disorder
  Manic-depressive/unspecified psychosis
  Anxiety states
  Phobic disorders
  Unspecified neurotic disorders
Parkinson's disease
Previous cancer
(Identified in NCI SEER Registry data
and recorded under Current Medical
  Renal failure
    Cerebrovascular accident
    Transient ischemic heart attack
  Thyroid or glandular disorders
  Urinary tract problems
    Chronic cystitis
  Other serious comorbidity
    Protein-calories malnutrition
    Postinflammatory pulmonary fibrosis
    Systemic lupus erythematosus
    Systemic sclerosis

Total Comorbidity

The number of separate comorbid conditions for each patient was totaled. For example, a patient with arthritis and urinary tract infection received a value of two, as did a patient with COPD and angina, because these conditions are considered distinct from each other. However, a patient with angina and cardiovascular disease received a value of one because both are "heart-related conditions."

Life Threat

Potential impact of the comorbid conditions on short term survival to the individual was assessed using a conceptual model, "Life Threat." Severity levels of the comorbidity burden were assigned based on comorbid conditions and their impact codes determined by the "current" and "history of" comorbid disease classifications. An algorithm of "Life Threat" including levels of high, moderate, low, and negligible was determined a priori based on clinical judgment, the literature on chronic diseases, and data on mortality.7-10

Disease Stage

SEER collects information on extent of disease, which was converted into disease stage based on the tumor, lymph node, and metastases (TNM) system specified by the American Joint Committee on Cancer.11


Information on date of death and cause of death for our patient sample was obtained through methods used routinely for follow-up of patients in the NCI SEER Program.1 Active follow-up is performed for all patients alive at last contact. A study cutoff date of December 31, 1993 was used to ascertain early mortality.

Statistical Procedures

The Mantel-Haenszel (M-H) test for trend was used to assess the ordinal relationship between two categoric variables.12 Category scores, required for the M-H test, were assigned in equal increments. Box and whisker plots displayed the distribution of total comorbidity by age group.13 Survival curves were estimated using the Kaplan-Meier method.14 Proportional hazards regression was used to explore the impact of combinations of explanatory variables on predicting survival from all causes of death.15 Logistic regression was used to identify variables associated with unknown stage of disease at diagnosis.16 Statistical analyses were performed using the SAS system.17 In these regression models, all explanatory variables (i.e., Total Comorbidity, Life Threat, comorbid conditions) were dichotomous. Total comorbidity was grouped by quintile. Patients whose disease stage was unknown were included in the survival analyses using an indicator variable. All P values were based on testing two-sided hypotheses. Backward selection procedures used the 0.05 level of significance for removing an explanatory variable from the model.


Characteristics of the Study Sample

The distribution of age, stage at diagnosis, Total Comorbidity, and Life Threat are summarized in Table 2 for women and men. The sampling protocol resulted in fairly equal numbers of male and female patients in each of the three age groups.

Table 2. Patient Characteristics by Gender
Patient characteristicsMale (n = 799)Female (n = 811)
  • a

    Patients sampled by age stratification.

  • b

    Mantel trend test, P = 0.03.

Age (yrs)a    
Stage at diagnosisb    
Total Comorbidities    
Life Threat    

Stage at diagnosis of colon carcinoma was available for 97% of the patients for both genders.

Only 3% of the colon carcinoma patients (47 patients) were diagnosed with Unknown Stage disease. A significant trend toward more unknowns in higher age groups (P = 0.04) was observed. More than half the patients were 75 years and older. Female patients tended to have a higher stage of disease at diagnosis than male patients (P = 0.03). This finding was not consistent across age groups. Total comorbidity was similar for men and women, ranging from no comorbid disease present (in 78 patients) to a high of 12-14 conditions recorded (in 6 patients). Forty percent of male patients and 41% of female patients had five or more conditions.

No significant trend toward High Impact Life Threat was noted for men compared with women. Greater than 80% of the patients of each gender experienced ≥ 1 comorbid conditions of High or Moderate Impact Life Threat. Stage distributions for men and women by age group are presented in Figure 2 (6K).

Figure 2.

Stage distributions for colon carcinoma patients by age and gender. Data are from the National Institute on Aging/National Cancer Institute Study, 1997. □: women; ▪: men.

The percentage of patients with unknown stage was highest in the age group ≥ 75 years. However, there was not a significant relationship between higher stage and older age group. The trend toward higher stage for women than men, observed in the study sample as a whole, was statistically significant in only the 65-74-years age group.

The relationship between Total Comorbidity (the number of comorbidities) and age group was striking (P < 0.0001). The median number of comorbidities increased with advancing age for both male and female patients (Fig. 3 (6K)). Within age strata, distributions were similar for both genders.

Figure 3.

Percentile distribution of total comorbidity for colon carcinoma patient by age and gender. Data are from the National Institute on Aging/National Cancer Institute Study, 1997. The top and bottom of the box represent the 75th percentile and 25th percentile of the distribution, respectively. The line in the middle of the box indicates the median. The top "whisker" extends to the 90th percentile; the bottom "whisker" to the 10th percentile.

The hierarchical distribution of the impact levels of Life Threat: (High, 4; Moderate, 3; Low, 2; Negligible, 1; and None, 0) by age and gender is shown in Figure 4 (10K).

Figure 4.

Percent distribution of Life Threat by age and gender for colon carcinoma patients. Data are from the National Institute on Aging/National Cancer Institute Study, 1997. equation image : No Comorbidity; equation image : Negligible Impact; □: Moderate Impact; equation image : Low Impact; ▪: High Impact.

Older persons vary considerably in manifestation of comorbidity. In each age group, all Life Threat levels were represented. Percentage rates varied from < 5% for the No Comorbidity category in the oldest age group to > 50% for the High Impact category in that same age group. A dramatic escalation in the percentage of men and women with High Impact Life Threat (Level 4) was noted as age group increased (with slightly higher rates for men); corresponding decreases in the percentages of Life Threat Impact were observed in the categories of No Comorbidity (Level 0), Negligible (Level 1), and Low (Level 2). The proportion of patients, male and female, with Moderate Life Threat comorbid conditions remained fairly stable across age groups.

Specific Comorbid Conditions

Anemia, hypertension, high severity heart problems, gastrointestinal (GI) problems, and moderate severity heart problems affected ≥ 20% of the study population in the High and Moderate Impact Life Threat categories; COPD, smoking, and, to a lesser extent, insulin-dependent diabetes were each cited in ≥ 5% of the patients. Low Impact Life Threat comorbidities affecting ≥ 10% were arthritis, urinary tract problems other than renal failure, and low severity heart problems.

The relationship between age group and presence of comorbid disease of High or Moderate Impact Life Threat or other comorbidities of current medical concern affecting at least 15% of the study population is shown in Table 3.

Table 3. Frequency and Percent of Men and Women with Selected Comorbidities by Age Group
 55-64 yrs65-74 yrs≥ 75 yrsTotal n = 1610
 Men n = 254Women n = 271Men n = 280Women n = 264Men n = 265Women n = 276
  1. HTN: hypertension; COPD: chronic obstructive pulmonary disease

  2. P <0.001, mantel trend test for age group, stratified by gender.

  3. P <0.005, mantel trend test for gender, stratified by age group.

Anemia70  (27.6%)75  (27.7%)83  (29.6%)96  (36.4%)124  (46.8%)144  (52.2%)592  (36.8%)
HTNa,b55  (21.7%)83  (30.6%)92  (32.9%)107  (40.5%)94  (35.5%)121  (43.8%)552  (34.3%)
Heart: High Impacta42  (16.5%)31  (11.4%)75  (26.8%)55  (20.8%)105  (39.6%)106  (38.4%)414  (25.7%)
Heart: Moderate Impacta31  (12.2%)35  (12.9%)66  (23.6%)59  (22.3%)85  (32.1%)64  (23.2%)340  (21.1%)
Heart: Low Impacta26  (10.2%)21  (7.7%)54  (19.3%)34  (12.9%)55  (20.8%)56  (20.3%)246  (15.3%)
Gastrointestinala49  (19.3%)41  (15.1)66  (23.6%)60  (22.7%)71  (26.8%)71  (25.7%)358  (22.2%)
Arthritisa,b24  (9.4%)37  (13.7%)47  (16.8%)59  (22.3%)53  (20.0%)78  (28.3%)298  (18.5%)
COPDa,b33  (13.0%)20  (7.4%)60  (21.4%)47  (17.8%)64  (24.2%)37  (13.4%)261  (16.2%)

The trend of increased prevalence by successively older age groups was significant for all comorbidities (P < 0.05 for GI problems in men and COPD in women; P < 0.005 for all others). Overall trends with age group, adjusted for gender, were significant (P < 0.0005).

Little difference in comorbidity prevalence was noted between male and female colon carcinoma patients. Anemia, a comorbid condition associated with colon carcinoma, was recorded as > 40% for both genders in the group age ≥ 75 years. Comorbidities with significant male/female differences are: hypertension in the youngest and oldest age groups; arthritis in the oldest age group only; and COPD in the youngest and oldest age groups (P = 0.05).

Early Mortality

Twenty-eight percent of the patients (n = 454) had died by the end of 1993. Disease stage at diagnosis was a major factor in survival. Even with a maximum follow-up of only 24 months, a significant trend toward worse survival was observed with increasing disease stage. Survival curves by disease stage are shown in Figure 5 (8K).

Figure 5.

Overall survival of colon carcinoma patients by age and disease stage at diagnosis (total/deaths). Data are from the National Institute on Aging/National Cancer Institute Study, 1997.

Greater than 50% of the deceased patients had Stage IV disease (n = 212) or Unknown Stage disease (n = 30). Colon carcinoma accounted for 73% of deaths (n = 330). Four percent of deaths (n = 17) were due to other cancers. Heart disease accounted for 9% of deaths (n = 41). Other major causes of death were COPD, vascular disease, and pulmonary emboli (n = 23; 5%). The cause of death for 3% of patients (n = 13) was unknown. The causes of death for the remaining 7% of the deceased patients in the sample (n = 30) were attributed to a variety of other conditions (e.g., pneumonia, infections, other pulmonary problems, GI hemorrhage).

Regression models were used to evaluate the impact of comorbidity on overall survival. Age group, gender, and stage were included in all models. Risk ratios and 95% confidence intervals for disease Stages II, III, IV, and Unknown Stage were compared with Stage I. As shown in Table 4A, there was an increasing risk of death with increasing stage. Patients with unknown stage were at a tenfold higher risk of death.

Table 4. Models of Overall Survival: Risk Ratios and 95% Confidence Intervals
  1. COPD: chronic obstructive pulmonary disease.

A. Relationship of overall survival to stage, adjusted for age group and gender
Stage categoriesRisk Ratio95% confidence interval
  Stage Ireference
  Stage II1.80[1.24, 2.62]
  Stage III2.41[1.65, 3.50]
  Stage IV11.98[8.53, 16.84]
  Unknown stage9.99[6.21, 16.08]
B. Relationship of overall survival to total comorbidity, adjusted for age group, gender, and stage
Total comorbidity categories:Risk ratio95% confidence interval
  31.33[0.98, 1.80]
  40.96[0.68, 1.36]
  5-61.44[1.10, 1.90]
  7-141.85[1.39, 2.46]
C. Relationship of overall survival to specific comorbidities Each comorbidity was tested in a model that included age group, gender, and stage against the reference category in which the specified comorbidity was not present. The results of the comorbidity comparison in 11 separate models are shown.
Comorbidity (present vs. not present)Risk ratio95% confidence interval
  Heart: high impact1.48[1.21, 1.82]
  COPD1.67[1.34, 2.08]
  Renal failure1.99[1.21, 3.29]
  Alcohol abuse2.20[1.23, 3.93]
  Deep vein thrombosis2.06[1.23, 3.47]
  Liver disease3.04[1.74, 5.30]
  Other serious comorbidity2.33[1.53, 3.55]
  Anemia1.25[1.04, 1.51]
  Depression1.63[1.12, 2.38]
  Diabetes1.37[1.05, 1.79]
  Thyroid/glandular1.49[1.09, 2.03]
D. Backward selection of comorbidities in Panel C. The inclusion of age group, gender, and stage was required. Each comorbidity was compared with the reference category in which the specified comorbidity was not present.
Included in all modelsRisk ratio95% confidence interval
  Age 55-64 yrs(reference) 
  Age 65-74 yrs1.35[1.05, 1.74]
  Age ≥ 75 yrs1.93[1.51, 2.47]
  Female1.20[0.99, 1.45]
  Stage I(reference) 
  Stage II1.86[1.27, 2.71]
  Stage III2.52[1.73, 3.68]
  Stage IV12.48[8.85, 17.58]
  Unknown stage9.47[5.86, 15.29]
Comorbidities retained (present vs. not present)  
  Heart: High Impact1.31[1.06, 1.61]
  COPD1.51[1.20, 1.89]
  Renal failure1.73[1.04, 2.89]
  Liver disease2.54[1.44, 4.49]
  Other serious comorbidity1.72[1.10, 2.67]
  Thyroid/glandular1.44[1.05, 1.98]

In Table 4B, Total Comorbidity added significantly to a model predicting survival (P = 0.0007). Risk ratios for Total Comorbidity were significantly different for the two highest quintiles (5-6 and 7-14) compared with the lowest (0-2). The model fit was not significantly improved when a comorbidity score that was weighted by Life Threat was used.

In Table 4C, several comorbidities considered individually added significantly to the model containing age group, gender, and disease stage. The interaction of comorbidity and stage did not qualitatively change the risk ratios. Interaction terms did not add significantly to the model for diabetes, renal failure, deep vein thrombosis, liver disease, other serious comorbidities, and depression (P > 0.05). There were insufficient data to evaluate the interaction terms for alcohol abuse. For the remaining significant comorbidities, the effect of adding the interactions in general was to increase the excess risk due to the comorbidity for lower stages and decrease the effect for higher stages. Except for two categories in thyroid and glandular disease (Stage III and Unknown Stage), there was always an added risk due to the comorbidity, regardless of whether interaction terms were present.

In Table 4D, backward selection applied to a model containing the comorbidities of Table 4C (age group, gender, and disease stage) retained High Impact heart problems, COPD, renal failure, liver disease, other serious comorbidities, and thyroid/glandular disease.


Ninety-three percent (n = 1494) of the patients received surgery. When the tumor is confined within the large bowel, surgery is potentially curative. Even in patients with advanced stage colon carcinoma, surgery is often performed to prevent bowel obstruction, to reduce symptoms, or is directed at the malignancy in the case of isolated metastases.

Of the 116 patients who did not have surgery, 55.2% (n = 64) were diagnosed with Stage IV disease. For 28.4% (n = 33), disease stage was unknown (for 23 of these patients, surgery was not recommended, suggesting that the colon carcinoma may have been in an advanced stage or that the patients' health status precluded surgery). The remaining 16.3% of patients (n = 19) who did not receive surgery had been diagnosed with Stages I and II disease. Surgical intervention was not recommended or contraindicated for 13 patients, refused by 1 patient, and not performed for unknown reasons in 5 patients.


The NIA/NCI SEER Study generated information on comorbidity by augmenting tumor registry data to help us understand more about the cumulative effects of aging and cancer and to ascertain the role comorbidity plays in increasing the risk for early mortality. Cancer in the elderly does not occur in isolation from other chronic health conditions. More than a decade ago, Dr. Alvan Feinstein observed that clinicians analyze the cancers, but not the patients.18 This comment also applies to epidemiologic research because such studies tend to focus on specific diseases (e.g., cancer) even though the morphology of the cancer and the results of antineoplastic therapy occur within the context of other health problems.

Dr. Feinstein's assertions were underscored in our earlier analyses (R.Y., L.A.G.R., and J.W.Y.), in which we used NCI SEER tumor registry data.19, 20 We found differences in treatment related to age, but could only speculate, in the absence of information on comorbidity, that intercurrent disease and/or the multiple pathology generally associated with aging accounted for these findings.

Our first research question asks what is the comorbidity burden of older colon carcinoma patients? Preexisting chronic conditions identified in the NIA/NCI SEER Study patient sample were found to be a combination of extremely serious conditions, possible precursors to major illnesses, and nonfatal conditions. The prevalence of many comorbidities increases with advancing age group for both genders and high percentages for heart-related conditions, anemia, hypertension, GI problems, and arthritis were observed. Although the impact of certain major comorbid conditions for High and Moderate Life Threat are somewhat more prominent in men (e.g., heart-related problems and COPD), and other comorbid conditions are slightly more prominent in women (e.g., anemia and hypertension), these differences are small compared with the increases in prevalence by age group. The pattern of prevalence percentage rates between male and female patients are remarkably similar. However, our data coincide with information reported in studies using national health data that men tend to have more serious chronic conditions whereas women tend to have more of the nonfatal chronic conditions.21-24

Often, survey reports emphasize contrasts between the number and types of conditions in men and women using national data, highlighting discrepancies in the incidence and prevalence of conditions according to gender.7, 8, 25 Reports based on these data are embedded in the context of life expectancy and seek explanations as to why women live longer than men.21, 23, 24 Interpretation of our results follow the thoughts of Dr. Lois Verbrugge: "Combing for sex differences can make us ignore the most basic conclusion, that the list of leading conditions is really very similar for the sexes. What bothers one sex bothers the other as well, even though the specific ranks may differ...and the pace of mortality may vary."21

Percentage levels of chronic conditions reported for the NIA/NCI patient sample for all three age groups, although analogous to those described in national surveys documenting disease characteristics for men and women, are higher than in national samples. The comorbidity burden reported for a hospitalized patient sample diagnosed with colon carcinoma is not directly comparable to health studies of the aged by community survey methods. Patients are evaluated and treated (for a very serious disease in this instance), leading to a more thorough ascertainment of comorbidity. Community-dwelling persons, presumed in reasonably good health, may have latent comorbid conditions that are unrecognized, not detected, or simply not recalled.

Methodologic studies on agreement between questionnaire data and medical records data show that health interview survey data underestimate the prevalence of chronic conditions in the population.26 Data obtained through self- and proxy reports are often inaccurate.27 Checklists of chronic conditions, variations in phrasing of questions, and using single informants for an entire household could lead to misreporting. However, with respect to the latter, proxy reports may not only be an important source of information regarding the elderly respondent, it may be the only source under certain circumstances. Certainly, a mixed mode approach that combines features of several methods is optimal. Survey field work with an elderly population requires modifications in procedures.26-28

The NIA/NCI SEER Study data show an unusually high prevalence of anemia and GI conditions in comparison with national data. Anemia due to the presence of internal bleeding (i.e., blood in the stool, rectal bleeding) and GI discomfort and pain (i.e., ranging from vague symptoms to bowel obstruction, tenesmus, and mucous diarrhea) are commonly associated with colon carcinoma. Patients with extensive disease also may have hematuria due to bladder invasion, and for female patients, bleeding or discharge due to vaginal invasion may be present as well.

The second research question asks if the comorbidity burden of colon carcinoma correlates with disease stage. No age/stage relationship for this malignancy was noted for our colon carcinoma patient sample. This finding is consistent with the literature reporting on age/stage relationships for colon carcinoma.29-31 Under a heavy burden of comorbid conditions, elderly patients may have less tendency to focus on cancer symptoms, delay attending to them, or the warning signs of cancer may be obscured by other medical problems (i.e., masked symptomatology not apparent to patients or physicians). In the case of the elderly, colon disturbance is often regarded as concomitant with old age.

Greater than 50% of the colon carcinoma patients of both genders were diagnosed with Stages I and II disease, and slightly less than 25% of the patients had Stage III disease. The remaining patients had Stage IV and Unknown Stage disease. Why disease stage was unknown may be attributable to several reasons: the stage was known, but not recorded in the medical record; the patient was not subjected to the procedure used for definitive staging (i.e., surgery) because there were obvious signs that the malignancy was far advanced; the severity of other pathologic conditions of the patient diagnosed with colon carcinoma precluded assessment of the tumor stage and there was limited probability of treatment success; consent to have surgery was not given; or any combination of these.

Our third research question addresses the influence of the comorbidity burden on survival. The total number of comorbidities and certain types of chronic conditions, most of which fall into the High Impact Life Threat category, increase the risk of early mortality for colon carcinoma patients. This pattern suggests that patients who have a greater number of conditions also are likely to be those suffering from a greater comorbidity burden. Some conditions were those associated with organ dysfunction due to metastases or acute clinical conditions that indicate impending death (e.g., liver "disease," renal failure, and those included in the category of other serious comorbidities). Other diseases (e.g., heart problems, COPD, thyroid/glandular problems) had a significant impact on survival outcome.

Concurrent diseases in elderly patients with colon carcinoma present clinical challenges in implementing therapeutic options. Although the clinician must take into account all risk factors involved in therapy, surgical intervention is often crucial to prevent and relieve bowel obstruction and anemia secondary to chronic blood loss. Accurate staging resulting from the surgery and abdominal exploration will lead to appropriate adjuvant therapy or supportive care. All aspects must be used to form a complete clinical picture, rather than simply equating age with fragility. Surgical complications may increase perioperative mortality in patients with certain comorbid conditions. These risks must be weighed against the possibility of cure or significant palliation. For example, we assume that because surgery was not performed for > 50% of the colon carcinoma patients with Stage IV disease and close to 75% of those with Unknown Stage, the fragility of these patients, their comorbid disease burden, other geriatric syndromes, the logistics of obtaining treatment, or the ability to give informed consent entered into this decision.

In general, risks of surgery for older patients are not different than for younger patients given the presence and severity of intercurrent disease. The surgical literature is replete with reports that address these issues with an admonition that has been succinctly summarized by Dr. W. Bradford Patterson "with careful patient selection and attention to comorbid conditions during pre-, intra- and postoperative periods, few surgical procedures need to be ruled out.  .  ."32

Our study has some limitations and several strengths. The approach using supplementary information to augment population-based registry data did not lend itself to application of risk prediction methods developed for in-hospital mortality and longitudinal studies developed by others.9, 33-35 Therefore, we designated a priori the spectrum of chronic conditions that were likely to increase patient vulnerability for death based on clinical judgment, the literature, and national mortality statistics.7-10 Another limitation is that the source of patient information, retrospective medical records review, was confined to documenting the patient's medical history and current problems available from hospital records. This method precluded obtaining information on pretreatment and posttreatment evaluation of older-aged patients and measures of physical performance of patients prior to or after the hospitalization episode.

Conversely, although refined evaluation of clinical factors was outside the scope of this study, the strength of our approach in acquiring information on comorbidity (derived from patient data entered in medical records of the cancer patients with confirmed diagnoses of colon carcinoma and survival follow-up) is more clinically accurate than could be obtained from patients' self-reports on their health status or from administrative data bases. These data are obtained from skillful probing by physicians' history and physical workups, laboratory tests, nursing notes, and other highly relevant information sources. Incidence and mortality data are validated in a quality-controlled tumor registry system. Thus, we were able to take advantage of a defined geographic population to attain a sample of a large number of older-aged patients. We utilized computer linkage between quality-controlled medical record abstracts and tumor registry reports to gain a better understanding regarding the comorbidity burden in the older patient "host."


The NIA/NCI SEER Study describes the high prevalence and nature of the comorbidity of older colon carcinoma patients and its interaction with the malignancy as a risk factor for early mortality. Population-based evidence support the importance of comorbidity and its consequences in the cancer course. Data also indicate that older age and the presence of comorbidity are predictive of not receiving surgery.

Ultimately, of all factors affecting the prognosis of patients with colon carcinoma, the single most important factor is early diagnosis for patients of all ages. Negative aspects of aging (i.e., comorbid conditions) were found to interact with colon carcinoma to affect early mortality. Diagnosis and treatment of a tumor are likely to induce great physical and emotional stress in a body already rife with other conditions, some of which have created marginal functioning in other organs or systems.36

The NIA/NCI SEER Study data definitively point to the urgent need for special clinical research efforts to provide cancer treatment guidelines for disease management in the presence of multiple pathology. In general, persons aged ≥ 65 years are underrepresented in prospective clinical trials.37 Concerns about the generalizability from clinical trials research to the older-aged population, the age segment primarily affected by cancer, should be raised.38-43 How can the evaluation of therapeutic options in clinical trials research, cancer control research, and epidemiologic investigations be made more pertinent to the majority of patients afflicted with colon carcinoma (i.e., the elderly)?

We shall continue to investigate the role of comorbidity in cancer outcome for other tumors in the elderly to gain a better understanding of its influence on outcome. The NIA/NCI SEER study data stimulate the promotion of cancer control research efforts to benefit older persons afflicted with cancer.


The authors would like to acknowledge the excellent working relationship and alliance established among the participants of the National Institute on Aging (NIA)/National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) Study on Cancer in the Elderly; the NCI SEER Principal Investigators study coordinators, and abstractors; the Information Management Services (IMS) staff; and the NIA/NCI leadership. They also thank Winifred K. Rossi, NIA, for her excellent editing assistance and preparation of the tables for this article.