Prognostic factors in 291 patients with pulmonary metastases submitted to thoracotomy Department of Pelvic Surgery-Hospital

Pulmonary metastases are among the most common sites for the spread of cancer, particularly with mesenchymal tumors. Despite improvements in cancer therapy, the prognosis remains poor, except for the highly selective group of patients who are considered eligible for surgical treatment and those with chemosensitive tumors. The main objective of this study was to evaluate survival results of 291 patients who underwent thoractomy due to pulmonary metastases between 1953 and 1986. The number of metastases ranged from a minimum of 1 to as many as 30. The type of resection depended on site, size, and number of pulmonary lesions: 154 wedge resections, 49 lobectomies, and 9 pneumectomies. The tumor was not resectable in the remaining 79 patients. Of the prognostic factors analyzed, only type of pulmonary resection (wedge vs. lobectomy vs. pneumectomy vs. unresectable) and disease free interval (DFI) between primary treatment and pulmonary metastases diagnosis «=6 months vs. 7-12 months vs. >12 months) were selected as independent predictors of the risk of death in multivariate analysis. A reduced model for bone tumors included disease free interval, sex and histology.


INTRODUCTION O
nly a small number of cases that underwent pulmonary metastases resection before 1950 were published, and most of those were lacking detailed information on survi val rates and prognostic factors.In 1884, Kronlein (13)  metastases resection from a chest wall sarcoma that was operated simultaneously.In 1926, Divis (8) performed a metachronous pulmonary metastases resection.In 1939, Barney and Churchill (3) resected one pulmonary metastases from a kidney carcinoma.This patient survived 23 years after nefrectomy.In 1947, Alexander and Haight (1) reported good sl;lrvival results in 3 of 6 patients that underwent pulmonary metastases resection.
Only in the 50's and 60's with the development of surgical and anesthetic techniques did pulmonary resections begin to be performed with more frequency.In spite of the fact that the number of published cases now exceeds 2,000, various controversies about indications and prognostic factors remain (12).
Some important indication factors have to be considered ( 16): (I) The primary tumor must be under control or able to be controlled; (II) pulmonary metastases resection must be technically possible; (III) the patient must not have another site of metastatic lesions; (IV) surgery should be considered only in the absence of a more efficient therapy; (V) the patient must have adequate clinical conditions and pulmonary function to be operated on.
The main objectives of this paper were to evaluate survival results and identify prognostic factors related to pulmonary metastases resection.These results can be used to improve accuracy when evaluating cases eligible for this kind of treatment.

CASEWORK AND METHODOLOGY
The authors reviewed the records of 291 patients who underwent pulmonary metastases resection at the Department of Pelvic Surgery of A. C. Camargo Hospital, Foundation Antonio Prudente, in Sao Paulo, Brazil, from January 1953 to December 1986.These patients represent only a portion of a larger group with clinical diagnosis of pulmonary metastases.Only patients eligible for surgical treatment were considered for this study.The criteria for the indication of surgical treatment was rather uniform during the period of this investigation, specifically: good performance status, no evidence of pleural effusion or mediastinum invasion and absence of a more efficient therapy.Patients were operated on independent of number, size, and location of pulmonary metastases.The histological type of primary tumor was not taken into consideration for treatment.When there was any plausible doubt whether or not the pulmonary lesion was in fact metastatic, the case was excluded from this study.
All cases were evaluated with chest X-rays.Chest planigraphy was done in 157 cases (55%).There were unilateral metastases in 259 cases (89%), and 32 ( The histology of the metastases was the same as the primary tumor in 247 (84.9%), different in 23 (7.9%), not conclusive in 10 (3.4%), and in 11 (3.8%) there was no information on the chart.
Statistical Analysis: product limit estimates of the survival function were used for the computation of cumulative survival rates (11), using the KMSURV program (5).Overall survival was computed on the basis ofthe dates and events abstracted from the original patient records.This was defined as the interval between the date of the thoractomy and the date of death for uncensored observations.
Cox's regression model was used to estimate the hazard ratio (HR) of death due to the combined effects of two or more study factors (4).This model was also used to build models containing the most restricted subset of variables with independent predictive properties with respect to the risk of death.A stepwise forward algorithm was used for variable selection.Inferences were based on the partial likelihood ratio (deviance) statistic between nested models (10% significance level).Finally, risk groups were determined on the basis of distribution of combined effects of significant variables, and split into three levels: low risk (RR<0.5 I), intermediate risk (0.51<RR<1.51)and high risk (RR>1.5 I).

RESULTS
The number of pulmonary metastases ranged from a minimum of I (102 cases, 35%) to as many as 30.The exact number of metastases was not reported in 9 cases, (3%).At the closing date of the study, there was a total of 6,193 patient-months of cumulative follow-up experience with a 10 month median.During the follow-up period, 74 patients (25.4%) were re-operated with further thoractomies: a second thoractomy in 58 cases and a third in 16 cases.One hundred and ninety-six (196) patients died (67.4%) due to residual or recurrent cancer or treatment complications (5.8% post-operative mortality rate 30 days after first thoractomy) and 5 deaths (1.7%) due to causes not related to cancer.Nineteen (19) patients were alive (6.5%) with recent follow-up information (1) OFI = disease free interval available (three of them with recurrent disease).An additional 76 patients (26.1 %) were considered lost to follow-up, but contributed sufficient information before then to be included in the survival analysis.
Table 1 shows a summary of cumulative survival data for selected characteristics.Demographic variables had no impact on prognosis.Disease free interval (DFI) prior to the detection of pulmonary metastases was significantly associated with patient's outcome.Patients with DFls longer than 6 and 12 months experienced better survival rates, (15.2% and 23.8% respectively for five year overall survival).The site of the primary tumor did not have significant effect on survival results.The number of metastases and treatment-related factors had a significant impact on survival rates.Rates were better for patients with solitary metastases and those who underwent wedge resections.The positive prognostic effect of pre-and/or post-operative chemotherapy was apparently substantial.However, its indication was remarkably biased by patients performance status and supposed chemosensitivity to the tumor.Therefore this variable was omitted from the pool of candidate prognostic factors used to build the Cox models for survival prediction purposes.
Patients with chondrosarcoma and bone tumors other than osteosarcoma had the highest survival rates in this series, with 50% of 5-year overall survival rates.On the other hand, the worst results were seen among primary liver cancer and malignant melanoma patients, both groups having no survivors after five years.(Table 2).
Only two of the prognostic factors were selected in the statistical model as independent predictors of the risk of death: the type of pulmonary resection and DFI.The model for bone tumors included DFI, sex and histology.The risk of death due to soft parts tumors was only affected by the number of metastases and DFI (Table 3).
A synthetic variable representing the risk of death was calculated using regression coefficients from the multivariate models shown in Table 3. Patients were subsequently grouped into three empirical risk groups (low, intermediate and high) based on the distribution of the synthetic linear predictors.The cumulative survival distribution for each of these risk groups is shown in figure 1.

DISCUSSION
The impact of various prognostic factors on survival rates of patients who underwent pulmonary metastases resection has been recognized in the literature.
Though DFI has been shown by various authors (6,7,9, I7,23,24,25,26,27) specifically to be the most important prognostic factor, other investigations have not confirmed this fact (10,18,2 I).In this study, however, DFI emerged as one of the most significant prognostic factors.
Another significant predictor of outcome is the type of pulmonary resection, a variable based on site, size and number of metastases (19,26).In summary, the larger the size of pulmonary metastases, the worse the prognosis, but on the other hand, total curative pneumectomy is still defended by several authors (6,9,14,16,17).Our results indicate that the type of resection was the most important predictor of survival.Patients submitted to wedge resections had a better prognosis.
Other papers dealing with squamous carcinoma of the head and neck (15,21,25) do not confirm those results.In this study the number of resected pulmonary metastases was a significant prognostic factor when solitary lesions were compared with two and three or more.However, patients with good performance status can reach the long term survival category in spite of multiple pulmonary metastases.In our opinion, the total number of metastases should not be considered as an absolute limiting factor for surgery, except when wide pleural dissemination occurs.The same is valid for recurrent lesions, when diagnosed by computerized tomography.
Chemotherapy was a significant prognostic factor in preliminary analysis; however we must bear in mind other possibly confusing variables such as the variety of types of primary tumors and various strategies of chemotherapy used over the 33 year range of the.study.In the case of osteosarcoma, for example, chemotherapy has been the first option as a neo-adjuvant therapy for pulmonary metastases for the last 20 years (15,19).Neo-adjuvant and adjuvant chemotherapy have also been used on Wilms' tumors, ovary and testicular carcinoma (2,7).The patients classified as having unresectable pulmonary disease in this study who survived more than 5 years had, in fact, chemosensitive tumors.Histology and site of primary tumor do not show any significant effect on survival results after pulmonary metastases resection.However, tumors like chondrosarcoma, kidney adenocarcinoma, Wilms' tumor, colorectal carcinoma and uterine cervix carcinoma tended to have a better prognosis.On the other hand, there were no survivors of malignant skin melanoma or primary liver cancer.
Our study confirmed that sex and age do not have a significant effect on survival results (10,20).
In conclusion, patients who have a DFI that is greater than 6 months, and a single lesion that is resectable with wedge surgery have the best prognosis.Histology and site of primary tumor must be analyzed case by case, but we think that surgery is still the only way to offer a last chance of prolonged and comfortable survival for many patients.
presented the first case of pulmonary Address for correspondence: Benedito M. Rossi Hospital A. C. Camargo -Funda9ao Antonio Prudente Rua Professor Antonio Prudente, 211 Sao Paulo -SP -Brasil-CEP 01509-010

Figure 1 -
Figure 1 -Cumulative survival distribution for low, intermediate and high risk of death.

Table 2
Cumulative survival rates according to histopatology of primary tumor

Table 3 Statistical models for the risk of death after surgical treatment of pulmonary metastases
• \\ \ j , .~. \,.