|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2002 American Society for Clinical Oncology Development and Validation of a Prognostic Model to Predict the Length of Survival in Patients With Carcinomas of an Unknown Primary SiteByFrom the Centre Régional de Lutte Contre le Cancer Val d’Aurelle, Montpellier; Institut Gustave Roussy, Villejuif; Institut Claudius Régaud, Toulouse; Centre Eugène Marquis, Rennes; Centre Paul Papin, Angers; and Centre Hospitalier Universitaire, Saint Etienne, France. Address reprint requests to Stéphane Culine, MD, PhD, Department of Medicine, Centre Régional de Lutte Contre le Cancer Val d’Aurelle, 34298 Montpellier Cedex 5, France; email: stculine{at}valdorel.fnclcc.fr
PURPOSE: To identify clinical and biologic variables with significant impact on survival in patients with carcinomas of an unknown primary site and to develop a simple prognostic model for the selection of patients in prospective clinical trials. PATIENTS AND METHODS: Univariate and multivariate prognostic factor analyses were conducted in a population of 150 unselected patients and led to the construction of two successive classification schemes. An external data set of 116 patients enrolled onto two prospective trials was used for validation. RESULTS: When studying clinical variables only, poor performance status (2 or 3) and presence of liver metastases were retained in the multivariate analysis. The first classification scheme consisted of three subgroups of patients with median survivals of 10.8, 6.0, and 2.4 months, according to the number of adverse prognostic factors. With the introduction of serum lactate dehydrogenase (LDH) levels in a further step, liver metastases were no longer significant. The second classification scheme therefore included poor performance status (relative risk [RR], 2.1) and elevated serum LDH level (RR, 2.1). Good-risk and poor-risk patients were identified, with median survivals of 11.7 months and 3.9 months, respectively (P < .0001). The 1-year survival rates were 45% and 11%, respectively. This second classification scheme was validated in an external data set: the median survival rates of patients assigned to the good-risk group and the poor-risk group were 12 months and 7 months, respectively (P = .0089). The 1-year survival rates were 53% and 23%, respectively. CONCLUSION: A simple prognostic model using performance status and serum LDH levels was developed and validated. It allows the assignment of patients into two subgroups with divergent outcome. Further prospective trials will be designed using this prognostic model.
CARCINOMAS OF AN unknown primary site (CUP) represent a group of heterogeneous tumors that share the unique clinical characteristic of metastatic epithelial disease with no identifiable origin at the time of therapy. The prognosis is generally poor, with a median survival of approximately 6 to 12 months.1 In recent years, significant advances have consisted almost exclusively of the identification of subsets of patients with clinical and pathologic features requiring specific guidelines that may translate into prolonged survival: women with axillary lymph nodes containing adenocarcinoma, primary papillary serous peritoneal adenocarcinomatoses, cervical lymph node metastases from squamous-cell carcinoma, middle-line undifferentiated carcinomas in young males, and undifferentiated carcinomas with neuroendocrine features.1 Unfortunately, the majority of CUP (approximately 85%) do not fall into one of these rather favorable subsets. Furthermore, the benefit of chemotherapy over best supportive care is still unknown and the optimal chemotherapy remains to be determined.2 We report here on the results of a prognostic factor analysis conducted in a population of 150 patients with CUP excluding clinical and pathologic subsets requiring specific treatments as described above. The results were validated in an independent set of patients. We specifically investigated the baseline characteristics that may be of prognostic value for survival with the aim to define prognostic subgroups of patients and then to design future prospective clinical trials dealing with chemotherapy according to these prognostic subgroups.
The medical and pathologic records of 150 patients with CUP who were consecutively registered and followed at the Montpellier Cancer Center from January 1989 to December 1999 were retrospectively reviewed. Approximately one third of them were included onto prospective clinical trials. Because of the great variability in clinical presentation, diagnostic procedures aiming at the discovery of a primary tumor site varied. For the purposes of the study, patients were required to undergo at least the following procedures: thorough history and physical examination, chemistry profile (including the determination of the following serum tumor markers in men: prostate-specific antigen, alpha-fetoprotein, and human chorionic gonadotrophin), chest radiographs, computed tomography scan of the chest and abdomen, and directed radiologic work-up of any symptomatic areas. A specific pathologic evaluation was required at diagnosis to confirm the epithelial origin of the disease and to exclude other malignancies and specific primary tumor sites. Patients were also excluded from the present study if they had any of the following features requiring well-defined treatments: women with adenocarcinoma that involved only axillary lymph nodes, women with primary papillary serous carcinoma of the peritoneum, patients with squamous carcinoma or neuroendocrine histology, and patients with carcinoma that involved a single potentially resectable tumor site. Survival was evaluated according to the Kaplan-Meier method and differences between the survival curves were assessed by the log-rank test.3,4 Independent prognostic factors were identified by the Cox regression analysis according to a backward stepwise selection procedure on all variables by removing nonsignificant variables in order of decreasing importance. Statistical analysis was performed using STATA software (STATA Corp, College Station, TX). The clinical charts of patients who entered the external data set for validation of the prognostic model were collected from two phase II prospective clinical trials using cisplatin-based regimens. The first study was a monocenter trial led at Institut Gustave Roussy (Villejuif, France) from 1993 to 1998 in 48 patients,5 and the second study was a multicenter trial performed by the French Study Group on Carcinomas of Unknown Primary from 1999 to 2001 in 78 patients.6
Patient Characteristics The initial characteristics of 150 patients are listed in Table 1. The median age was 57 years (range, 16 to 84 years). Well-differentiated adenocarcinoma was the most common histologic diagnosis. Approximately one third of patients had only one site of metastatic disease. Bone, lung, and liver were the dominant visceral sites of disease, whereas mediastinum was the most frequent location of lymph node involvement. One third of patients had a low performance status (2 or 3). A platinum-based regimen was delivered in 121 (86%) of 140 patients who were treated with chemotherapy. The median survival was 7.5 months. The 1-year and 2-year survival rates were 30% and 11%, respectively. At the time of analysis, 140 of 150 patients had died.
Prognostic Model With Clinical Variables Only The results of the univariate analysis are listed in Table 2. Survival was significantly related to the following pretreatment adverse prognostic factors: a poor performance status (2 or 3, P < .0001), the presence of liver metastases (P < .006), and more than two metastatic sites (P = .07). In the multivariate analysis, two independent adverse prognostic parameters were retained: a poor performance status and the presence of liver metastases (Table 3). These results were used to derive a classification scheme based on the number of adverse prognostic factors (Table 4). Seventy-two patients (48%) with no adverse prognostic factor (good-risk group) had a median survival of 10.8 months and a 43% 1-year survival. Eighteen patients (12%) with both adverse prognostic factors had a median survival of 2.4 months and a 0% 1-year survival. Regarding the outcome of patients with only one adverse prognostic factor, the median survival was 7.0 months in 29 patients with liver metastases and 4.1 months in 31 patients with a poor performance status. Because these survival rates were not statistically different (P = .11), we combined both subgroups. Therefore, 60 patients (40%) with one adverse prognostic factor (poor performance status or liver metastases, intermediate-risk group) had a median survival of 6.0 months and a 23% 1-year survival.
Prognostic Model Including Both Clinical Variables and Biologic Parameters The serum levels of a number of biologic parameters (normal v elevated) obtained before chemotherapy had started were studied in the second step. In univariate analysis (Table 5), four biologic parameters had prognostic relevance: alkaline phosphatase (P = .004), carcinoembryonic antigen (P = .07), CA-125 (P = .07), and lactate dehydrogenases (LDH) (P = .0001). In multivariate analysis, survival was shown to be independently influenced by serum LDH levels when adjusted for performance status. As shown in Table 6, liver metastases had no independent significant value. LDH levels were correlated with the presence of liver metastases. For the 81 patients with known LDH levels, only four (8%) of 52 with no liver metastasis had an elevated LDH value. Among the 29 patients with liver metastases, 15 (52%) had normal LDH levels. On the basis of the observation that 34 of 35 patients with a good performance status (0 or 1) and no liver metastasis had a normal serum LDH value, patients were assigned to the good-risk group if they had a good performance status and a normal serum LDH level (n = 44) or a good performance status and no liver metastasis (n = 37). Therefore, in the second prognostic model (Table 5), the good-risk group included 81 patients (59%), with a median survival of 11.7 months, and 57 patients (41%) were assigned to the poor-risk group, with a median survival of 3.9 months (P < .0001) (Fig 1). The 1-year survival rates were 45% and 11%, respectively (Table 7).
Validation of the Second Prognostic Model The second classification scheme was validated in an external data set of patients included on two successive prospective trials in independent cancer centers. Full data for performance status and serum LDH levels were available in 116 (92%) of 126 patients. As compared with the reference population, there were significantly more patients assigned to the poor-risk group (58% v 41%) because of a higher proportion of patients with elevated serum LDH levels (54% v 22%). Conversely, there were fewer patients with a poor performance status (15% v 33%). The median survival rates of patients assigned to the good-risk group and the poor-risk group were 12 months and 7 months, respectively (P = .0089) (Fig 2). The 1-year survival rates were 53% and 23%, respectively.
The design of treatment plans for patients presenting with CUP remains a daily challenge for physicians. Except for subgroups of patients with clinical and pathologic features requiring specific guidelines that may translate into prolonged survival, the benefit of chemotherapy remains questionable, although such a benefit is suggested by historical comparison.7 Because conclusions drawn from small series are limited, many authors have advocated for the design of randomized trials in CUP patients.8,9 For example, investigators from the Sarah Cannon Cancer Center recently reported two prospective phase II studies with markedly different response rates and survival results using carboplatin- and paclitaxel-based regimens.10,11 As the authors acknowledged themselves, the discrepancies in the results are more likely related to the inclusion of a higher proportion of patients with poor prognostic features in the later study rather than to a difference in anticancer activity of these rather similar regimens. In this context, the ability to adequately assign patients into prognostic subgroups using a simple, reliable classification scheme predicting the length of survival would be useful in the design of clinical trials. Previous studies dealing with prognostic features in patients with CUP identified a number of independent adverse variables such as male sex, a poor performance status, a high number of metastatic sites, the presence of liver metastases, and an elevated serum alkaline phosphatase level (Table 8).12-15 Using performance status and serum alkaline phosphatase level as independent prognostic factors, Van der Gaast et al built a prognostic model in a population of 79 patients.15 Three subgroups were identified with median survivals of more than 4 years, 10 months, and 4 months according to the presence of zero, one, or two adverse prognosticators, respectively. However, this model was limited to CUP patients with poorly differentiated adenocarcinoma or undifferentiated carcinomas and was not validated in an independent data set of patients. Investigators at the M.D. Anderson Cancer Center published a regression tree analysis in 1,000 consecutive patients that led to the identification of 10 subgroups with median survivals ranging from 5 to 40 months.16 However, the applicability of this tree analysis in daily practice (and even in clinical trials) is not convenient. Therefore, no simple, reliable prognostic model had been reported so far for the management and design of clinical trials in CUP patients.
The aims of the present study were the development and validation of such a prognostic model for clinical research. The goal was achieved through two successive steps. In the first multivariate analysis with clinical variables only, a poor performance status and the presence of liver metastases were identified as adverse prognostic variables. The number of metastatic sites was only retained in the univariate analysis. The male sex and the histologic type had no significant prognostic relevance. In the first classification scheme, patients were assigned into three subgroups according to their performance status and the presence or not of liver metastases. This approach did not seem to be satisfactory because the differences between median survivals (10.8, 6.0, and 2.4 months for patients with good, intermediate, and poor prognosis, respectively) did not attain sufficient significant clinical impact. However, this scheme was shown, first, to explain our approach in developing a reliable and accurate model and, second, to confirm in our reference set the results observed in previous experiences when liver metastases were shown to have independent prognostic value.12,14,16 When incorporating biologic parameters into the second step, serum LDH levels were shown to influence survival in multivariate analysis. The second classification scheme was therefore based on performance status and serum LDH levels because the presence of liver metastases was no longer significant. Two subsets of patients with a clearly different expected median survival (11.7 and 3.9 months for patients with good and poor prognosis, respectively) were identified. This model was validated in an external set of patients. It is noteworthy that the median survival in the poor-risk group ranged from 3.9 months in the reference population to 7 months in the validation set. In our opinion, this discrepancy was expected because of the selection of patients: patients were unselected in the reference population, whereas all patients included in the validation set were enrolled onto clinical trials. The consequence was that the proportion of patients with the most powerful adverse prognostic factor (ie, a poor performance status) was lower in the validation set. In the past, we have explored several strategies of chemotherapy in patients with CUP in both monocentric and multicenter prospective trials: high-dose chemotherapy with hematopoietic stem-cell support17; dose-dense chemotherapy with hematopoietic growth factor support18; chemotherapy based on histologic differentiation5; and screening of new regimens, namely, cisplatin in combination with gemcitabine or irinotecan.6 Further prospective trials led by our national group will be designed using the prognostic model described in the present study. In good-risk patients, the end point of our next trial will be the determination of the optimal chemotherapy regimen. In patients with poor-risk disease, low-toxicity chemotherapy will be challenged with best supportive care only.
1. Hainsworth JD, Greco FA: Treatment of patients with cancer of an unknown primary site. N Engl J Med 329: 257-263, 1993 2. Sporn JR, Greenberg BR: Empirical chemotherapy for adenocarcinoma of unknown primary tumor site. Semin Oncol 20: 261-267, 1993[Medline] 3. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53: 457-481, 1958[CrossRef] 4. Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 50: 163-167, 1966[Medline]
5. Saghatchian M, Fizazi K, Borel C, et al: Carcinoma of unknown primary site: A chemotherapy strategy based on histological differentiation—Results of a prospective study. Ann Oncol 12: 535-540, 2001 6. Lortholary A, Culine S, Bouzy J, et al: Cisplatin in combination with either gemcitabine (GC) or irinotecan (IC) in carcinomas of unknown primary (CUP): Results of a randomized phase II study. Proc Am Soc Clin Oncol 21: 153a, 2002 (abstr 609) 7. Daugaard G: Unknown primary tumours. Cancer Treat Rev 20: 119-147, 1994[CrossRef][Medline] 8. Rigg A, Cunningham D, Gore M, et al: A phase I/II study of leucovorin, carboplatin and 5-fluorouracil (LCF) in patients with carcinoma of unknown primary site or advanced oesophagogastric/pancreatic carcinoma. Br J Cancer 75: 101-105, 1997[Medline] 9. Greco FA, Gray J, Burris HA, et al: Taxane-based chemotherapy for patients with carcinoma of unknown primary site. Cancer J 7: 203-212, 2001[Medline] 10. Hainsworth JD, Erland JB, Kalman LA, et al: Carcinoma of unknown primary site: Treatment with 1-hour paclitaxel, carboplatin, and extended-schedule etoposide. J Clin Oncol 15: 1651-1656, 1997
11. Greco FA, Burris HA, Litchy S, et al: Gemcitabine, carboplatin, and paclitaxel for patients with carcinoma of unknown primary site: A Minnie Pearl Cancer Research Network study. J Clin Oncol 20: 1651-1656, 2002 12. Kambhu SA, Kelsen DP, Fiore J, et al: Metastatic adenocarcinomas of unknown primary site: Prognostic variables and treatment results. Am J Clin Oncol 13: 55-60, 1990[Medline] 13. Hainsworth JD, Johnson DH, Greco FA: Cisplatin-based combination chemotherapy in the treatment of poorly differentiated carcinoma and poorly differentiated adenocarcinoma of unknown primary site: Results of a 12-year experience. J Clin Oncol 10: 912-922, 1992[Abstract]
14. Abbruzzese JL, Abbruzzese MC, Hess KR, et al: Unknown primary carcinoma: Natural history and prognostic factors in 657 patients. J Clin Oncol 12: 1272-1280, 1994
15. Van der Gaast A, Verweij J, Planting AST, et al: Simple prognostic model to predict survival in patients with undifferentiated carcinoma of unknown primary site. J Clin Oncol 13: 1720-1725, 1995
16. Hess KR, Abbruzzese MC, Lenzi R, et al: Classification and regression tree analysis of 1000 consecutive patients with unknown primary carcinoma. Clin Cancer Res 5: 3403-3410, 1999
17. Culine S, Fabbro M, Ychou Y, et al: Chemotherapy in carcinomas of unknown primary site: A high-dose intensity policy. Ann Oncol 10: 569-575, 1999 18. Culine S, Fabbro M, Ychou Y, et al: Alternative bimonthly cycles of doxorubicin, cyclophosphamide, and etoposide, cisplatin with hematopoietic growth factor support in patients with carcinoma of unknown primary site. Cancer 94: 840-846, 2002[CrossRef][Medline] Submitted April 1, 2002; accepted August 12, 2002.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2002 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
