The influence of comorbidities, age, and performance status on the prognosis and treatment of patients with metastatic carcinomas of unknown primary site

A population-based study

Authors


Abstract

BACKGROUND

The authors investigated how comorbidities, age, and performance status were related to the choice of chemotherapy and to the prognosis of patients with carcinomas of unknown primary site (CUP).

METHODS

Patients in Northern Alberta who were diagnosed with CUP during 2000 to 2003 were included (n = 389 patients). Survival was compared by age at diagnosis (ages <65 years, 65-74 years, and >75 years), comorbidity score (Adult Comorbidity Evaluation-27 [ACE-27] scores of 0-1 and >2), performance status (PS), and other explanatory variables, such as gender, histology, and site and number of metastases.

RESULTS

The median age was 68 years, and the median overall survival was 12 weeks. An ACE-27 overall comorbidity score >2 was found in 34% of patients, and a PS >2 was observed in 50% of patients. Multivariate analysis showed that patients who had a PS ≥2 and a high overall ACE-27 score had a worse prognosis. The impact of comorbidities on survival was limited to patients with low PS. Patients who were not evaluated at a cancer center were older, had a worse functional status, and had more moderate or severe comorbidities. Among the 257 patients who were evaluated at a cancer center, 108 patients received chemotherapy, and 121 patients had a good PS (0-1). Age was the only independent variable that was related to the likelihood of not receiving chemotherapy among patients who had a good PS. The median overall survival of the 121 patients who had a good PS was 317 days, and overall survival was not associated significantly with chemotherapy. A logistic regression analysis that included all patients who were evaluated at a cancer center identified young age, good PS, lymph node/pleural involvement, and few comorbidities as variables that were associated independently with receiving chemotherapy.

CONCLUSIONS

Patients with CUP who were not evaluated at a cancer center were older, had a worse functional status, and had more moderate or severe comorbidities; this referral bias largely explained the differences between data from registries and from tertiary centers. Moderate and severe comorbidities impacted survival in patients with who had a PS ≥2. An age-related decline was observed in the percentage of adults with good PS who received chemotherapy. The current results suggested that older patients with CUP were under treated and that factors other than PS were involved in the decision to use chemotherapy for the treatment of patients with CUP. Cancer 2006. © 2006 American Cancer Society.

Cancer of unknown primary site (CUP) represents a group of heterogeneous tumors that share the unique clinical characteristic of metastatic epithelial disease with no identifiable origin at the time of diagnosis, and it comprises from 2% to 5% of all new cancer diagnoses.1 The prognosis for patients with untreated CUP is very poor, with a median survival of 3 to 4 months, thereby strengthening arguments for a nihilist approach.2 In the late 1980s, Greco et al. demonstrated that patients with poorly differentiated carcinoma and “extragonadal syndrome” features benefited from platinum therapy.3 Since then, several other clinicopathologic entities have been identified that confer a better prognosis with treatment.1, 4 These subsets of patients include women with axillary lymph node metastasis, peritoneal adenocarcinomatosis, cervical lymph nodes that contain squamous carcinoma, and carcinomas with neuroendocrine features. More recently, it was shown that the use of taxanes (paclitaxel or docetaxel) in combination with a platinum compound produced improved outcomes for patients who do not fit into these favorable subsets.5, 6 Those nonrandomized clinical studies showed promising results with a median survival that ranged from 8 months to 13 months and with 1-year survival rates from 15% to 29%. These lengths and rates of survival are similar to those observed for patients with chemotherapy-treated metastatic lung, pancreatic, and gastric carcinomas, among others.7 In a review of >600 patients with CUP who were evaluated for treatment at The University of Texas M. D. Anderson Cancer Center (M. D. Anderson; a tertiary referral center) from 1987 to 1992,8 the median survival was 11 months, 90% of patients received anticancer treatment, and 49.2% of patients received chemotherapy.

In 2 recently published studies from population-based registries, CUP was assessed from 1984 to 1992 and from 1984 to 1993.9, 10 van de Wouw et al. reported on the incidence, histologic findings, survival, and prognostic factors among 1285 patients who were treated in general hospitals and were entered in the Eindhoven Cancer Registry.9 Those investigators observed a median survival of 11 weeks in the biopsy-proven group and a 1-year survival rate of 15%. Two-thirds of those patients received supportive measures only, and only 9% received chemotherapy. Compared with patients who were evaluated at M. D. Anderson, patients in the Dutch registry were 10 years older, whereas the histopathologic factors, the male:female ratio, and favorable prognostic factors were similar. However, important prognostic factors, such as performance status (PS) and lactate dehydrogenase levels (LDH), were not assessed in the registry studies.9, 10 Culine et al. recently developed a prognostic model that stressed the importance of PS and LDH in a population of patients with CUP who were evaluated at a French cancer center from which the previously described clinically and pathologically favorable subsets were excluded.11 Their results subsequently were validated in an independent data set of patients who were enrolled on 2 prospective trials.11 When studying clinical variables only, those investigators observed that poor PS and presence of liver metastasis were retained as significant variables in the multivariate analysis. Moreover, PS was the primary determinant of whether a patient with CUP would benefit from chemotherapy.4, 12 Patients with a PS <2 are considered good candidates, and patients with a PS >2 are treated with supportive care.

Others prognostic factors, including comorbidities, have never been formally assessed in patients with CUP, which may explain the discrepancy in outcomes between patients who are evaluated at tertiary centers and unselected patients who are captured by registry studies. Comorbidities are independent prognostic factors for patients with cancer.13 Correlations between the severity of comorbid conditions and outcomes have been observed among patients with head and neck cancer,14 breast cancer,15 prostate cancer,16 colon cancer,17 and endometrial cancer.18 The presence and severity of comorbidities are associated with age but are distinct from the overall PS.19 Several studies have demonstrated that comorbid conditions also may play an important role in decision making.15, 20, 21 To reconcile the different outcomes reported by registries and referral centers, we conducted a population-based clinical analysis of consecutive patients with CUP in the Northern Alberta Cancer Registry from 2000 to 2003 to investigate how comorbid conditions, PS, and age may influence the choice of chemotherapy and the prognosis of patients with CUP.

MATERIALS AND METHODS

The Northern Alberta Cancer Registry has data on all patients who are diagnosed with cancer in Northern Alberta, an area with 1.5 million inhabitants. In Alberta, a patient with cancer cannot receive chemotherapy at a noncancer center. The medical records of patients with CUP who were registered consecutively in the North Alberta Cancer Registry from January 1, 2000 to December 31, 2003 were reviewed retrospectively. Eligible patients were required to undergo at least the following procedures: thorough history and physical examination, determination of prostate-specific antigen (PSA) level, chest radiographs, pelvic ultrasound, and directed radiologic work-up of any symptomatic area. We excluded from further analysis those patients who met the following criteria: 1) a clinical diagnosis without histologic proof of malignancy, excluding patients with cerebral metastasis; 2) an obvious primary tumor identified at the time of their initial visit; 3) CUP of nonepithelial origin; and 4) patients who were registered inappropriately. Women with adenocarcinoma that involved only axillary lymph nodes (1 patient), women with primary peritoneal carcinoma (24 patients) patients with cervical lymph nodes that contained squamous carcinoma (22 patients), middle-line undifferentiated carcinomas in young men (3 patients), and patients with carcinomas that contained neuroendocrine features (38 patients) were excluded, because these subsets of patients require specific guideline-based therapy.

With the approval of the relevant institutional review boards and treating physicians, 1 researcher (P.S.) extracted additional information on patient, tumor, and treatment characteristics and clinical outcomes. The research database included information on age, gender, referral center, histology, date of diagnosis, location(s) of metastasis, number of metastasis, PS, LDH, type of surgery, chemotherapy, radiotherapy, hormone therapy, overall survival, and comorbidities.

Age at diagnosis was analyzed according to the following categories: younger than age 65 years, ages 65-74 years, and older than age 75 years. The choice of the 3 age-specific cohorts was based on the publications from the National Institute of Aging and the National Institutes of Health, which defined the elderly as patients older than age 65 years with subcategories of the “young-old” (ages 65-74 years) and “older-old” (older than age 75 years).22

Patient PS was extracted from the medical records by using the World Health Organization (WHO) scale. If this item was not recorded formally, then we estimated a patient's PS on the basis of the written medical records. Patients were grouped into 1 of 2 PS categories: PS 0 or 1 and PS 2, 3, or 4.

Comorbidities were assessed by using the Adult Comorbidity Evaluation 27 (ACE-27) at the time of diagnosis. The ACE-27 is a new, 27-item validated comorbidity index for use with cancer patients.13 The ACE-27 grades specific diseases and conditions into 1 of 4 levels of comorbidities: Grade 0 (no comorbidity), Grade 1 (mild comorbidities), Grade 2 (moderate comorbidities), or Grade 3 (severe comorbidities) The index may be viewed online (http://cancercomorbidities.wustl.edu/web_courses_support/comorbiditiescoding). The overall comorbidity score is established by the highest scoring illness, except when there are ≥2 Grade 2 ailments present, in which case, the Grade is 3. Data on comorbidities were extracted from hospital admissions, referral letters, discharge summaries, medical histories, and medication usage. Because of the limited number of patients in each comorbidity group, patients were grouped for comparison into 1 of 2 categories: none/mild comorbidities or moderate/severe comorbidities.

Statistical Analysis

Univariate and multivariate logistic regression analyses were used to determine the association and hazard ratios for the choice of chemotherapy treatment and the presence of comorbidities in patients with a PS of 0 or 1, adjusting for the explanatory variables (age, gender, were evaluated at the Cross Cancer Institute versus a community cancer center, overall comorbidities score [ACE-27], histology, PS, and sites of metastases). The explanatory variables were dichotomized, and univariate analysis was used to establish their significance for inclusion in the multivariate model. To ensure the optimum combination of explanatory variables, only variables with P values <.10 were included.

Univariate and multivariate Cox regression analyses were used to determine the association between overall survival and PS, comorbidities, and age adjusting for covariates, metastases, LDH, and the number of metastatic sites. The univariate analysis was carried out to determine the significance of each predictor for overall survival; then, the multivariate analyses were performed by using a backward, stepwise selection process to derive a multivariate model of significant predictors (significance was prespecified as P <.05).

Survival was defined as the time from pathologic diagnosis until death or the end of the study (if the patient still was alive on January 1, 2005). The Kaplan–Meier method was used to generate overall survival curves, and comparisons between groups based on PS status, comorbidities, and age were made by using the log-rank test.

Because poorly differentiated or undifferentiated carcinomas have been considered more sensitive to chemotherapy, we compared these histologic subsets with others. All statistical analyses were performed using SPSS software (version 13.0; SPSS Inc., Chicago, IL).

RESULTS

Patient and Tumor Characteristics

Three hundred eighty-nine patients met the entry criteria for this study (Table 1). One hundred ninety-five patients were male (50%). The mean age was 68 years (range, 19-94 years). Patient, tumor, and treatment data are shown in Table 1. In this cohort, there were 149 patients (38%) younger than age 65 years, 117 patients (30%) ages 65-74 years, and 123 patients (32%) older than age 75 years. The distribution of comorbid ailment scoring revealed 16% severe (Grade 3), 18% moderate (Grade 2), and 27% mild (Grade 1) comorbidities; whereas 29% of the population had no comorbidity, and comorbidities were unknown for 10%. Fifty percent of patients had a low PS (2, 3, or 4). Almost 50% of patients received a specific treatment. A platinum-based regimen was received by 102 of 108 patients (94%) who were treated with chemotherapy. Two hundred twenty patients were evaluated at the Cross Cancer Institute (56.5%), 37 patients were evaluated at associated cancer centers (9.5%), and 132 patients were not evaluated (34%). Among the patients who were not evaluated at a cancer center, we were not able to assess PS in 33% of patients or comorbidities in 25% of patients. At the time of the current analysis, 344 of 389 patients had died. The median survival was 84 days (range, 0-1,524 days). The 1-year and 2-year survival rates were 19% and 4%, respectively.

Table 1. Patient, Tumor, and Treatment Data from 389 Patients with Metastatic Carcinoma of Unknown Primary Site
CharacteristicsNo. of Patients (%)
  • ACE-27: the Adult Comorbidity Evaluation 27 scale.

  • *

    Others included 5 undifferentiated tumors, 5 mixed histologic tumors, and 13 cerebral metastasis without histologic proof of cancer.

Age 
 Range, y19–94
 Median, y68
 <65 y149 (38)
 65–74 y117 (30)
 >75 y123 (32)
Gender 
 Female194 (50)
 Male195 (50)
ACE-27 overall comorbidity score 
 Grade 0112 (29)
 Grade 1105 (27)
 Grade 262 (18)
 Grade 371 (16)
 Unknown39 (10)
Histology 
 Well differentiated adenocarcinoma194 (50)
 Poorly differentiated adenocarcinoma or carcinoma119 (31)
 Undifferentiated carcinoma31 (8)
 Squamous cell carcinoma22 (6)
 Others*23 (6)
Performance status 
 037 (10)
 1101 (26)
 285 (22)
 391 (23)
 427 (7)
 Unknown48 (12)
Site of disease 
 Brain32 (8)
 Lymph nodes155 (40)
 Lung82 (21)
 Pleura76 (20)
 Liver153 (39)
 Peritoneum90 (23)
 Adrenal glands24 (6)
 Bone75 (19)
 Cutaneous nodes15 (4)
 Others66 (17)
No. of metastatic sites 
 1168 (43)
 2128 (33)
 >293 (24)
Lactate dehydrogenase level 
 <Normal80 (21)
 >Normal102 (26)
 Unknown207 (53)
Treatment 
 Chemotherapy88 (23)
 Radiotherapy61 (16)
 Chemotherapy and radiotherapy20 (5)
 Hormone therapy4 (1)
 Others (surgery and chemoembolization)13 (3)
 None215 (55)
Evaluated 
 Cross Cancer Institute220 (56)
 Affiliated cancer centers37 (10)
 No132 (34)

Table 2 displays patient and tumor characteristics across the age cohorts and according to the ACE-27 comorbidity index. In the subgroup of patients older than age 65 years, a higher comorbidity score was observed. High comorbidity scores were associated with reduced PS (73.5% vs. 26.5% of patients with a PS <2; P <.0001) and male gender (61.8% vs. 38.2% for female gender; P = .001). Compared with patients who had a PS <2, patients who had a PS >2 had significantly more moderate or severe comorbidities of the cardiovascular system (24% vs. 9%; P = .000373), the respiratory system (7% vs. 1%; P = .013), and the endocrine system (10% vs. 4%; P = .02; data not shown).

Table 2. Patient Characteristics, Tumor Characteristics, and Treatment by Age and Comorbidity Score
CharacteristicNo. of Patients (%)
AgeACE-27 Comorbidity Score
<65 y>65 y> 75 yP0–12–3P
  • ACE-27: the Adult Comorbidity Evaluation 27 scale.

  • *

    Others included 5 undifferentiated tumors, 5 mixed histologic tumors, and 13 cerebral metastasis without histologic proof of cancer.

No. of patients149117123 217133 
Gender   .72  .001
 Female73 (49.0)54 (47.0)67 (54.5) 123 (56.7)50 (37.6) 
 Male76 (51.0)63 (53.0)56 (45.5) 94 (43.3)83 (62.4) 
ACE-27 overall comorbidity score   <.001  
 Grade 0–1106 (74.6)58 (54.7)53 (52.0)    
 Grade 2–336 (25.4)48 (45.8)49 (48.0)    
Histology   .13  .83
 Adenocarcinoma67 (45.0)59 (50.4)68 (55.3) 110 (50.7)69 (51.9) 
 Other subtype*82 (55.0)58 (49.6)55 (44.7) 107 (49.3)64 (48.1) 
Performance status   .57  <.001
 0–158 (42.3)44 (40.4)36 (37.9) 104 (49.8)32 (26.9) 
 2–479 (57.7)65 (59.6)59 (62.1) 105 (50.2)93 (73.1) 
No of metastatic sites   .18  .22
 158 (38.9)39 (33.3)71 (57.7) 85 (39.2)61 (45.9) 
 >291 (61.1)78 (66.7)52 (42.3) 132 (60.8)72 (54.1) 
Treatment   <.001  <.001
 Chemotherapy69 (46.3)30 (24.8)9 (9.8) 82 (43.2)26 (13.5) 
 Radiotherapy33 (22.1)33 (26.5)15 (13.8) 50 (26.8)31 (18) 
 Other7 (4.7)5 (1.7)5 (3.3) 12 (1.9)5 (1.5) 
 No56 (37.6)74 (63.2)95 (77.2) 79 (36.7)91 (68.4) 
Evaluated at a cancer center   <.001  <.001
 Yes116 (77.9)88 (75.2)53 (43.1) 178 (82.0)79 (59.4) 
 No33 (22.1)29 (24.8)70 (56.9) 39 (18.0)54 (40.6) 

Survival Analysis Results

In the univariate analysis, short survival was related significantly to the following pretreatment factors: age older than 65 years (vs. younger than 65 years; P = .006), PS >2 (vs. PS <2; P <.001), ACE-27 Grade >2 (vs. ACE-27 Grade <2; P <.001), adenocarcinoma histology (vs. other histologic subtypes; P = .004), presence of liver metastases (P <.001), presence of peritoneal metastases (P <.001), lack of lymph node metastases (P = .014), number of metastatic sites >2 (P <.001), and high serum LDH level (P <.001). The multivariate analysis of the same factors identified PS >2 (P <.001), presence of liver metastases (P = .002), presence of peritoneal metastases (P = .02), high comorbidity score (P = .027), and number of metastatic sites >2 (P = .049) as independently significant predictors of survival (Table 3). A PS >2 yielded a hazard ratio of 3.31 (95% confidence interval [95% CI], 2.34-4.70), and an overall comorbidity score >2 yielded a hazard ratio of 1.46, (95% CI, 1.04-2.05). Age, histology, the number of metastatic sites, and a lack of lymph node metastases were not prognostic factors. Because of missing data, the LDH level was not incorporated into the multivariate analysis.

Table 3. Multivariate Analysis of Overall Survival According to Patient Characteristics: Variables of Independent Significance
VariableOverall Survival
HR (95% CI)P
  1. HR: hazard ratio; 95% CI: 95% confidence interval; ACE-27: the Adult Comorbidity Evaluation 27 scale.

ACE-27 overall comorbidity score: 2–3 vs. 0–11.46 (1.04–2.05).027
Performance status: 2–4 vs. 0–13.31 (2.34–4.7)<.001
Liver metastases: Yes vs. no1.83 (1.36–2.25).002
Peritoneal metastases: Yes vs. no1.8 (1.10–2.96).020
No. of metastatic sites: >2 vs. 11.55 (1.0–2.4).049

Because of the correlation between a low PS and overall ACE-27 comorbidity scores >2, subsequently, we assessed the prognostic factor of comorbidities according to PS. In the subset of patients with a PS <2, a high comorbidity score was not correlated with overall survival, but it was the strongest, significant, independent variable among patients with a PS >2 (P <.001). Among the patients with a PS >2, an overall comorbidity score >2 yielded a hazard ratio of 1.84 (95% CI, 1.33-2.53). Both liver and peritoneal involvement and the number of metastases were correlated significantly with low survival in the 2 groups, whereas gender, age, histology, and other site involvement were not significant.

Patient Referral and Treatment

The 132 patients (34%) who were not evaluated at a cancer center had a median overall survival of 21 days; whereas, among the 257 patients who were evaluated at a cancer center, the median overall survival was 150 days (P <.001) (Fig. 1). Patients who were not evaluated at a cancer center were older (median age, 75 years vs. 63 years; P <.001), had a worse functional status (PS >2, 83.3% vs. 51%; P <.001), and had more moderate or severe comorbidities (overall ACE-27 comorbidity scores of 2 or 3, 58.1% vs. 30.7%; P <.001) (Table 4).

Figure 1.

These are overall survival (OS) curves for 257 patients who were evaluated at cancer center (median OS, 150 days) and for 132 patients who were not evaluated at a cancer center (median OS, 21 days; P <.001).

Table 4. Clinical Description of Evaluated and Chemotherapy-Treated Patients with Metastatic Carcinoma of Unknown Primary Site
VariableNo. of Patients (%)
Full PopulationEvaluated PopulationNonevaluated PopulationChemotherapy-Treated Patients
  • ACE-27 indicates the Adult Comorbidity Evaluation 27 scale.

  • *

    P <.001 for comparison with patients who were evaluated at a cancer center.

No. of patients389257132109
Age, y    
 Median68.063.075.0*59.5
 Range19–9419–9435–9319–78
Gender    
 Female194 (50.0)132 (51.4)62 (47.0)64 (59.3)
 Male195 (50.0)125 (48.6)70 (53.0)44 (40.7)
Performance status    
 0–1138 (40.5)123 (49.0)15 (16.7)*61 (56.5)
 2–4203 (59.5)128 (51.0)75 (83.3)47 (43.5)
Overall comorbidity score: ACE-27    
 Grade 0–1217 (62.0)178 (69.3)39 (41.9)*82 (75.9)
 Grade 2–3133 (38.0)79 (30.7)54 (58.1)26 (24.1)

Univariate and multivariate logistic analyses showed that age older than 75 years, moderate and severe comorbidities, peritoneal involvement, and a PS ≥2 were correlated significantly with the likelihood of not being evaluated at a cancer center (data not shown). In the multivariate analysis, age older than 75 years yielded a hazard ratio of 5.01 (95% CI, 2.68-9.36), an overall comorbidity score >2 yielded a hazard ratio of 2.21 (95% CI, 1.20-4.09), and a PS >2 yielded a hazard ratio of 3.92, (95% CI, 1.85-8.30).

Among the patients who were evaluated at a cancer center, 175 patients received at least 1 specific anticancer treatment (chemotherapy, radiotherapy, or hormone therapy). Among them, 108 patients received chemotherapy. Although guidelines suggest that chemotherapy should be reserved for patients who have a good PS, 47 of 108 chemotherapy-treated patients had a PS ≥2. Therefore, we analyzed factors that were associated with receiving chemotherapy in the evaluated population. Among the predefined variables, young age (65 years), female gender, none or mild comorbidities, a good PS, the presence of lymph node and pleural metastases, and the absence of lung metastases were significantly more frequent in treated patients. Multivariate regression analysis showed that young age, none or mild comorbidities, lymph node and pleural involvement, and a good PS were associated independently with the likelihood of receiving chemotherapy (Table 5).

Table 5. Multivariate Logistic Regression Analysis of the Factors Associated with Receiving Chemotherapy among 257 Patients who Were Evaluated at a Cancer Center
Variable*Odds Ratio (95% CI)Chi-Square Statistic
  • 95% CI: 95% confidence interval; ACE-27: the Adult Comorbidity Evaluation 27 scale.

  • *

    The reference categories were age >65 years, male gender, being evaluated at the Cross Cancer Institute, overall ACE-27 comorbidity score of 2–3, other histologic subtype, lack of lymph node metastases, lack of liver metastases, lack of lung metastases, lack of bone metastases, lack of pleural metastases, lack of peritoneal metastases, lack of brain metastases, and a performance status of 2–4.

Age <65 y vs. 65–74 y3.06 (1.62–5.78)0.0006
Age <65 y vs. >75 y6.41 (2.66–15.38)<0.0001
Female gender1.71 (0.96–3.08)0.07
ACE-27 overall comorbidity score 0–11.92 (1.02–3.62)0.04
Lymph node metastases1.98 (1.10–3.57)0.02
Pleura metastases2.39 (1.10–5.17)0.084
Brain metastases0.41 (0.12–1.46)0.17
Performance status 0–11.65 (1.21–2.25)0.001

Factors Associated with Not Receiving Chemotherapy among 121 Patients with Good PS

Among the 121 patients who had a PS of 0 or 1, 61 patients received chemotherapy. Among the 60 patients who did not received chemotherapy despite a good PS, 39 patients had only 1 site of organ involvement, 29 patients received radiotherapy, 7 patients underwent surgery, 1 patient received hormone therapy, 1 patient underwent chemoembolization, and 26 patients received only best supportive care. The proportion of patients with a single site of metastasis was greater in the group of patients with a good PS who did not receive chemotherapy than in the group of patients with a PS ≥2 who did not receive chemotherapy (39 of 60 patients vs. 31 of 81 patients, respectively; P = .001); however, the proportion of patients who received radiotherapy did not differ statistically within the 2 groups (29 of 60 patients vs. 32 of 81 patients, respectively).

The 121 patients who had a good PS had a median overall survival of 317 days (range, 4-1502 days). Survival did not differ significantly between patients who received or did not receive chemotherapy (median overall survival, 327 days vs. 315 days, respectively; P = .45). Among the predefined variables, age (65-75 years and older than 75 years), female gender, moderate and severe comorbidities, and the absence of lung metastases were significantly more frequent in untreated patients. The multivariate regression analysis showed that ages 65-74 years and age older than 75 years were associated independently with the likelihood of not receiving a specific treatment (Table 6). Among 55 patients younger than age 65 years, 14 patients did not receive any chemotherapy; whereas, among 42 patients ages 65-74 years, 28 patients did not receive chemotherapy; and, among patients older than age 75 years, 18 of 24 patients did not receive chemotherapy. Patients ages 65-74 years and patients older than age 75 years had the highest odds ratios for not receiving chemotherapy: 3.48 (95% CI, 1.34-9.01) and 7.04 (95% CI, 2.23-22.24), respectively.

Table 6. Multivariate Logistic Regression Analysis of the Factors Associated with Not Receiving Chemotherapy among 121 Patients with a Performance Status of 0 or 1
Variable*Odds Ratio (95% CI)Chi-Square Statistic
  • 95% CI: 95% confidence interval; ACE-27: the Adult Comorbidity Evaluation 27 scale.

  • *

    The reference categories were age <65 years, male gender, overall comorbidity score ACE-27 score 0–1, other histologic subtype, lack of lymph node metastases, lack of liver metastases, lack of lung metastases, lack of bone metastases, lack of pleura metastases, lack of peritoneal metastases, and lack of brain metastases.

Age  
65–74 y3.48 (1.34–9.00)0.01
>75 y7.04 (2.23–22.24)0.0009
Female0.55 (0.24–1.25)0.15
Overall comorbidity score: ACE-27 Grade 2–32.21 (0.78–6.19)0.13
Lung metastases0.38 (0.13–1.14)0.084

DISCUSSION

The current study was the first to out knowledge that assessed the impact of comorbidities on the prognosis and treatment of patients with CUP. Our primary objective was to identify the factors that may reconcile the highly discrepant reports of prognosis and treatment described by tertiary referral centers and registry data.8–11

We retrospectively reviewed clinical data fro, 389 patients who had a diagnosis of CUP in the Northern Alberta Cancer Registry from 2000 to 2003. The characteristics in our population were similar to those in other populations described by registries.9, 10 Compared with CUP described by tertiary centers,11 our patients were 10 years older, had a worse PS, received less treatment, and had a worse prognosis with a median overall survival of 11 weeks. For example, in the study by Culine et al., the median age was 57 years, only 33% of their patients had a low PS (2 or 3), >90% of their patients received chemotherapy, and the median survival was 7.5 months.11

Our findings confirmed the value of prognostic variables reported by both registry and tertiary referral centers and further validated the clinical prognostic model established by Culine et al., in which the PS and the presence of liver metastases were the most important prognostic factors.11 Moreover, our study also validated previous reports that peritoneal involvement,23 the number of metastases,8 and moderate or severe comorbidities were prognostic features in our unselected population.

We also assessed the correlations between survival, PS, and comorbidities and observed that a PS ≥2 was associated with a high number of comorbidities. However, comorbidities only carried independent prognostic value in patients with a PS >2. This finding is concordant with both the “competing-demand” and “physiologic” hypotheses. The physiologic hypothesis suggests that the type of comorbidities and cancer interact at the physiologic level to increase aggressiveness or metastasis.24 The competing-demand hypothesis suggests that the management of severe chronic disease may constitute a competing demand that diverts resources and attention from the delivery of an appropriate diagnostic work-up or that existing health conditions may cause physicians to overlook cancer symptoms.25 The excess of moderate or severe comorbidities in patients with a low PS was related to involvement of the cardiovascular system, respiratory system, and endocrine system (in patients with poorly controlled diabetes or with end-organ failure). However, it is unclear whether these diseases actually affect the aggressiveness of the disease, represent a barrier to diagnosis, or are a competing demand on the physician's time and focus. However, we noted that less treatment and less chemotherapy were given with increasing comorbidity.

We observed important differences between patients who were or were not evaluated at a cancer center. Nonevaluated patients were significantly older, had more severe comorbidities, and had a worse functional status. The prognosis was poorer in nonevaluated patients, with a 3-week median overall survival. These data support a strong ‘referral bias’ and explain the better outcomes reported in studies from referral cancer center populations compared with registry-based populations.

Among the 257 patients who were evaluated at a cancer center, 174 patients received a specific treatment, and 108 patients received chemotherapy. Although evidence-based medicine does not define a standard for the systemic treatment of a CUP site that does not belong to a specific anatomoclinical entity,4 chemotherapy is considered only for patients who have good general health status (WHO PS, 0 or 1).4, 12 Considering these data, we analyzed the factors that are associated with chemotherapy in this subgroup of patients. Among the 121 patients who had a PS of 0 or 1, 61 patients received chemotherapy. The multivariate regression analysis showed that only age was related significantly with the likelihood of not receiving chemotherapy. Several studies showed that more elderly patients with various forms of cancer often did not receive chemotherapy or received less extensive treatment than their younger counterparts.15, 18, 26–29 The proportion of untreated patients was particularly high among patients with metastatic solid tumors26 and was most notable among patients older than age 70 years15, 27 and older than age 75 years.18, 28, 29

Although guidelines suggest that chemotherapy should be reserved for patients with a good PS, 47 of 108 chemotherapy-treated patients in the current study had a PS ≥2. Therefore, we analyzed factors that were associated with receiving chemotherapy in the evaluated population. This multivariate regression analysis also showed that young age was related significantly to the likelihood of receiving chemotherapy. A good PS, as expected,4 was associated with receiving chemotherapy. However, few comorbidities and lymph node/pleural metastases were associated independently with receiving chemotherapy. Several studies showed that comorbid conditions may play an important role in decision-making among cancer patients.15, 20, 21 Chemotherapy is considered more effective in patients with predominantly lymph node/pleural metastases from CUP, whereas it offers only limited benefit to patients with liver, bone, or multiple organ involvement.6 These data suggest that the site of metastasis may influence the decision to administer chemotherapy. Our results suggest that many factors other than PS influence decision-making in patients with CUP.

It is noteworthy that we observed an overall median survival of 317 days among patients who had a good PS and who were evaluated at a cancer center, which approximated the survival reported in tertiary centers and in nonrandomized studies. This suggests that patients from tertiary care centers and patients who are enrolled in nonrandomized trials are selected and do not reflect the overall population of patients with CUP.

We note several limitations of our analysis. First, our retrospective study has a greater proportion of missing data in patients who were not evaluated at a cancer center. Second, although computed tomography scans of the abdomen are not considered standard for diagnostic evaluation,4 other authors have recommended them for the initial work-up.1, 30, 31 Systematic computed tomographic scanning may have changed the diagnosis in some patients. Third, we were unable to assess socioeconomic status and race, which reportedly influence the administration of chemotherapy in cancer patients.32 Other important factors, such as variations in physician and patient preferences, may constitute barriers to chemotherapy and were not assessed.

Despite these limitations, our study led to several conclusions. We confirmed for the first time the validity of the clinical prognostic model established by Culine et al. in unselected patients from a cancer registry. We observed that moderate and severe comorbidities had an impact on survival in patients with a PS ≥2. Our logistic regression analysis of patients with good PS showed that both “old-old” patients and “young-old” patients received less chemotherapy than younger patients. In this group of patients with good PS, survival did not differ significantly whether or not chemotherapy was administered. Logistic regression analysis of patients who were evaluated at a cancer center also showed that young patients received more chemotherapy. However, in addition to PS, lymph node/pleural involvement and no or mild comorbidities were associated independently with receiving chemotherapy. These data suggest that many patients with CUP were treated according to current guidelines and that patients from tertiary centers and who are enrolled in nonrandomized trials are a highly selected population with an expectation of a better prognosis. Therefore, our results underline the necessity of randomized controlled trials in patients with CUP to define more clearly the value of chemotherapy and the impact of clinical variables on the effect of chemotherapy.

Acknowledgements

The authors thank Mary Burns for her administrative assistance.

Ancillary