This study was conducted by the Eastern Cooperative Oncology Group (Robert L Comis, M.D., Group Chair). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
Eastern Cooperative Oncology Group (ECOG) Study E1594 compared paclitaxel and cisplatin with three newer chemotherapy doublets in the treatment of patients with advanced nonsmall cell lung carcinoma (NSCLC). The accrual of patients with an ECOG performance status (PS) of 2 was discontinued due to a perceived rate of unacceptable toxicity.
Patients were stratified by PS and randomized to one of the following treatments: 1) paclitaxel (135 mg/m2) over 24 hours with cisplatin (75 mg/m2) on a 21-day cycle; 2) cisplatin (100 mg/m2) with gemcitabine (1 g/m2) on Days 1, 8, and 15 on a 28-day cycle; 3) cisplatin (75 mg/m2) with docetaxel (75 mg/m2) on a 21-day cycle; and 4) paclitaxel (225 mg/m2) over 3 hours with carboplatin (area under the curve, 6). All tests of statistical significance were two-sided.
Sixty-eight patients with an ECOG PS of 2 were enrolled, and 64 patients were evaluable for toxicity and response. Fifty-six percent of 64 evaluable patients were male, and 81% had Stage IV disease. Grade 3–4 hematologic toxicities occurred in > 50% of the patients in each treatment group. Nonhematologic Grade 3–4 toxicities occurred significantly less often in the paclitaxel and carboplatin arm (P = 0.0032). The overall rate of toxicity did not differ significantly from the rate of toxicity in the PS-0 or PS-1 cohorts. There were 5 deaths (7.35%) among 68 patients with a PS of 2 during therapy; however, only 2 of those deaths were attributed to therapy. The overall response rate for the 64 evaluable patients was 14%. The overall median survival of all 68 patients with a PS of 2, as determined by an intent-to-treat analysis, was 4.1 months.
It is now well accepted that cisplatin-based chemotherapy confers an improvement in overall survival and offers substantial palliation for patients with metastatic nonsmall cell lung carcinoma (NSCLC) compared with best supportive care.1–3 In the late 1990s, multiple studies revealed that the addition of newer agents to cisplatin, including gemcitabine,4 vinorelbine,5 and paclitaxel,6 offered a survival advantage compared with cisplatin alone or cisplatin combined with an epipodophyllotoxin or vinca alkaloid.
However, these advances apparently are confined to patients with an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0–1.7, 8 Patients with a PS of 2 have not attained the same benefits: Traditionally, it has been believed that these patients have an increased predilection for toxicities from therapy, counterbalancing any gains that may be achieved. Retrospective reviews of prospective, randomized studies of cisplatin-based chemotherapy trials conducted by cooperative groups have shown clearly that patients with a PS of 2 have a poorer survival.7, 9 In the ECOG experience with 1960 evaluable patients from five trials exploring cisplatin-based chemotherapy, a PS of 2 was associated with a statistically lower median survival of 3.3 months compared with 6.4 months for patients with a PS of 1 and 9.4 months for patients with a PS of 0 (P < 0.05).8
ECOG study E5592 compared paclitaxel (24-hour infusion) plus cisplatin with etoposide plus cisplatin in patients with a PS of 0 or 1 and revealed a superior overall response of > 25% for the low-dose (135 mg/m2) and high-dose (250 mg/m2 with granulocyte-colony stimulating factor support) paclitaxel arms compared with 12.0% for the etoposide plus cisplatin arm (P < 0.001).6 A survival advantage was noted for patients in the paclitaxel arms; the 2-year overall survival rate was 14.1%, and the median survival was 9.9 months for patients in the low-dose and high-dose cohorts combined compared with 11.3% and 7.6 months, respectively, for patients in the etoposide-containing arm (P = 0.048). There was no significant difference between the low-dose and high-dose paclitaxel arms. The combination of cisplatin plus 135 mg/m2 paclitaxel given as a 24-hour infusion, therefore, was chosen as the standard arm for the next ECOG trial in patients with metastatic NSCLC (ECOG study E1594). That study compared cisplatin plus paclitaxel with three experimental arms, all of which included a platinum combined with a newer agent. All three experimental arms yielded promising response rates and survival outcomes in Phase II trials.10 The experimental doublets were cisplatin plus gemcitabine, cisplatin plus docetaxel, and carboplatin plus paclitaxel (3-hour infusion). Study E1594 opened in October 1996 and closed in May 1999. Given recent improvements in supportive care and the sense that newer regimens were less toxic and more effective, patients with a PS of 2 were considered eligible. This was a departure from the previous ECOG policy, which confined participation to patients with a PS of 0 or 1. However, this trial was designed to assess separately the toxicity in patients with a PS of 2 after the first 50 patients were enrolled. In March 1997, the Data Monitoring Committee observed a substantial rate of toxicity in the patients with a PS of 2 and recommended continued, close monitoring. By August 1997, 66 patients with a PS of 2 had been enrolled, and accrual was discontinued because of a perception of excess adverse events. The determination of excess adverse events was based on the rate of toxicities and the 5 of 66 patients (7.6%) who died on study. This report represents a final analysis of this subgroup.
MATERIALS AND METHODS
Eligibility stipulated histologically confirmed Stage IIIB (pleural effusion) or Stage IV NSCLC. Written, informed consent and Institutional Review Board approval were mandated, and patients were required to have an ECOG PS ≤ 2 and no prior systemic therapy. Patients with clinically stable brain metastases managed by surgery and/or radiation therapy were eligible. Adequate hematologic (absolute neutrophil count > 2000/mm3 and platelets > 100,000/mm3), renal function (creatinine < 1.5 mg/dL), and hepatic function (aspartate transaminase > 5 times the upper limit of normal and bilirubin < upper limit of normal) were required at entry.
Four treatment arms were employed (Table 1): 1) paclitaxel 135 mg/m2 over 24 hours followed by cisplatin 75 mg/m2 on Day 2 every 21 days (PC arm); 2) cisplatin 100 mg/m2 on Day 1 plus gemcitabine 1 g/m2 on Days 1, 8, and 15 on a 28-day cycle (GC arm); 3) cisplatin 75 mg/m2 on Day 1 plus docetaxel 75 mg/m2 on Day 1 of a 21-day cycle (DC arm); and 4) paclitaxel 225mg/m2 on Day 1 administered over 3 hours with carboplatin dosed with an area under the curve (AUC) of 6 (PCb arm). The premedication and hydration regimens varied according to the chemotherapy administered. For both arms that included paclitaxel, 20 mg of oral dexamethasone was given 12 hours and 6 hours prior to the paclitaxel infusion with intravenous cimetedine (300 mg) and intravenous diphenhydramine (50 mg) given 1 hour prior to the taxane infusion. Intravenous fluids with appropriate electrolyte support were administered before and after the cisplatin infusion. Cisplatin was administered in 500–1000 cc of intravenous fluids over 1 hour. Gemcitabine was given over 30 minutes in normal saline. For the 1-hour docetaxel infusion, 8 mg of oral dexamethasone were given every 12 hours for 5 doses starting 24 hours prior to the infusion. If the alkaline phosphatase was 5 times the upper limit of normal and the aspartate transaminase was 1.6–5.0 times the upper limit of normal, then the docetaxel dose was decreased to 55 mg/m2.
Table 1. Treatment Arms
P: paclitaxel; C: cisplatin; G: gemcitabine; D: docetaxel; Cb: carboplatin; AUC: area under the curve.
Paclitaxel 135 mg/m2 over 24 hours on day 1 and cisplatin 75 mg/m2 on Day 2
Every 21 days
Gemcitabine 1 g/m2 on Days 1, 8, and 15 and cisplatin 100 mg/m2 on Day 1
Every 28 days
Docetaxel 75 mg/m2 on Day 1 and cisplatin 75 mg/m2 on Day 1
Every 21 days
Paclitaxel 225 mg/m2 over 3 hours on Day 1 and carboplatin AUC 6 on Day 1
Every 21 days
Grading of toxicities was according to the National Cancer Institute Common Toxicity Criteria. An adverse event was defined as any untoward medical event that occurred from the first dose of study medication until 30 days after the final dose, regardless of whether it was considered related to treatment.
Definition of Response
A complete response (CR) was defined as the complete disappearance of all clinically detectable disease measured by physical examination and/or radiographic studies for a period of at least 4 weeks. A partial remission (PR) was defined as a decrease ≥ 50% in the sum of the products of the two longest perpendicular dimensions of all measurable lesions for a period of at least 4 weeks without an increase > 25% in the size of any area known to contain malignant disease and without the appearance of any new areas of malignancy. Progressive disease (PD) was defined as an increase ≥ 25% in the size of measurable lesions. Time to disease progression was measured from the time of initiation of therapy to the time PD was documented, and survival was measured from the initiation of therapy to the last clinic visit or death. Both the time to disease progression and overall survival were determined by an intent-to-treat analysis.
All three experimental arms were compared with the PC arm. All tests of statistical difference were two sided. The Fisher exact test was used to compare response rates, and the log-rank test was used to evaluate survival.
The median age for the entire cohort of 64 evaluable patients was 63.6 years (range, 42–79 years) (Table 2). The median ages were 60.6 years (range, 43–78 years) for patients in the PC arm, 68.5 years (range, 47–78 years) for patients in the GC arm, and 65.2 years (range, 42–79 years) and 61.6 years (range, 44–73 years) for patients in the DC and PCb arms, respectively. Gender distribution revealed that the PC and DC arms were comprised of 67% and 61% males, respectively, whereas the GC and PCb arms had 46% and 47% males, respectively (Table 2). More than 80% of all patients had Stage IV disease except for the GC arm, in which 69% of patients had Stage IV disease. Weight loss ≥ 10% was noted in 28% of patients in the PC arm, in 23% of patients in the GC arm, and in 22% and 33% of patients in the DC and PCb arms, respectively.
Toxicity data were available for 64 patients. One patient withdrew consent before randomization, another patient had a myocardial infarction after registration, and a third patient died before chemotherapy was administered. A fourth patient was entered after patients with a PS of 2 were no longer being accrued. Any Grade 3–4 toxicity in patients with a PS of 2 occurred in 30–60% of patients in the PC arm (n = 18 patients), 8–67% of patients in the GC arm (n = 13 patients), 12–59% of patients in the DC arm (n = 18 patients), and 27–33% of patients in the PCb arm (n = 15 patients). The only statistical difference in any of the toxicities between the four treatment arms was that patients in the PCb arm had an overall lower incidence of Grade 3–4 toxicities compared with patients in the other three arms (P = 0.0032). When the data are evaluated in tabular form (Tables 3 and 4), it can be seen that the lower toxicity in patients in the PCb arm was due to 1) a slightly lower overall incidence of Grade 4 neutropenia and 2) the absence of Grade 3 or 4 nausea and emesis or diarrhea compared with patients in the other arms. There was no appreciable difference in neuropathies between the groups.
Table 3. Grade ≥ 3 Hematologic Toxicities and Adverse Events: Maximum-Patient Specific Incidence
There was no statistical difference in the incidence of hematologic toxicities in patients in the PC arm compared with patients in any of the other arms (Table 3). Among patients in the PC arm (n = 18 patients), Grade 3 and 4 granulocytopenia occurred in 20% and 40% of patients, respectively, and one episode of grade 4 febrile neutropenia was reported. Twenty-five percent of patients had Grade 3 anemia. Grade 3 and 4 granulocytopenia was noted in 25% and 33% of patients, respectively, in the GC arm (n = 13 patients), and Grade 3 and 4 thrombocytopenia occurred in 25% and 25% of patients, respectively, and 33% of patients experienced Grade 3 anemia. Among the patients in the DC arm (n = 18 patients), Grade 3 and 4 neutropenia occurred in 24% and 35% of patients, respectively; and two patients (12%) experienced Grade 4 febrile neutropenia, including one death due to gram negative sepsis. One patient (6%) had Grade 3 anemia. There also was one fatal episode of hemoptysis despite a normal platelet count. In the PCb arm (n = 15 patients), Grade 3 and 4 neutropenia occurred in 20% and 27% of patients, respectively. Only one patient had Grade 3 thrombocytopenia, and 20% of patients experienced Grade 3 anemia.
In the PC arm (n = 18 patients), 33% of patients had Grade 3 nausea, and Grade 3 and 4 emesis occurred in 5% and 35% of patients, respectively (Table 4). Grade 3 neuropathy was reported in 15% of patients. One patient (6%) had Grade 3 renal impairment. In the GC arm (n = 13 patients), 42% of patients had Grade 4 emesis. Grade 3 and 4 renal impairment occurred in 8% and 16% of patients, respectively, with one patient succumbing to a fatal episode of renal failure that was attributed to therapy and commenced 2 days after the first cycle. Autopsy revealed features of a microangiopathic process, like what is seen in the hemolytic uremic syndrome, a phenomenon associated with gemcitabine.11 Grade 3 and 4 infection occurred in 8% of patients each. There were three reports (13%) of Grade 3 neuropathy. Another patient died of progressive mental obtundation and seizures. No radiographic studies or autopsies were performed to determine whether this was secondary to cerebral metastases. Of the 18 eligible patients with toxicity data on the DC arm, Grade 3 and 4 emesis occurred in 6% and 35% of patients, respectively, and Grade 3 neuropathy occurred in 18% of patients. Grade 4 diarrhea occurred in 18% of patients, and Grade 3 and 4 allergic reactions occurred in 6% of patients each. One patient with bronchoalveolar carcinoma had a cardiorespiratory arrest on the day anticoagulation was started for a deep venous thrombosis (DVT). Despite anticoagulation and intubation, the patient died of further respiratory insufficiency. On the PCb arm (n = 15 patients), 20% of patients experienced Grade 3 neuropathy, and there was no Grade 3 or 4 emesis.
Comparison with Patients with a PS of 0 or 1
No difference was seen in the incidence of worst grade toxicities in the patients with a PS of 2 compared with patients who had a PS of 0 or 1. The incidence rates of worst Grade 3–4 toxicity among patients with a PS of 0 or 1 were 20–68% in the PC arm (n = 295 patients); 20–69% in the GC arm (n = 290 patients); 22–61% in the DC arm (n % 295 patients); and 28–53% in the PCb arm (n = 289 patients).12 The incidence rates of death on study were 4% for patients in both the PC arm and the GC arm, 6% for patients in the DC arm. and 3% for patients in the PCb arm. A Fisher exact test showed that there was no statistical difference between the incidence of worse grade toxicity when analyzed by all patients who had a PS of 2 compared with patients separately who had a PS of either 0 or 1. Nor was a difference seen when the patients who had a PS of 2 were compared with all patients who had a PS of 0 or 1 combined (P = 0.126). There was no statistical difference in the incidence of worst grade toxicity in patients with a PS of 2 in any of the specific arms compared with the incidence in patients with a PS of 0 or 1 separately or combined in the corresponding arm. It is noteworthy that 5 of 68 patients (7.35%) with a PS of 2 died on study. Table 5 shows that only two deaths (3%) were due to drug toxicities. This incidence rate is not appreciably different from the 4% Grade 5 toxicity rate observed in the patients with a PS of 0 or 1.
Table 5. Fatal Events
G: gemcitabine; C: cisplatin; D: docetaxel; DVT: deep venous thrombosis.
There was no significant difference between cohorts.
Response rate (no.)
Time to progression (months) (no.)
Median survival (months)
Overall 1-yr survival (%)
Of the 68 patients with a PS of 2 who were enrolled, 64 patients were evaluable for response (Table 5). The overall response rate was 14% (95%confidence interval [95%CI], 5.6–22.6%). The response rates for each arm did not differ significantly between arms: PC arm, 3 of 18 patients (17%; 95%CI, 3.4–39.6%); GC arm, 3 of 13 patients (23%; 95%CI, 5.5–57.2%); DC arm, 1 of 18 patients (6%; 95%CI, 1.0–27.3%); and PCb arm, 2 of 15 patients (13%; 95%CI, 1.7–40.5%).
Time to Progression
The overall median time to progression (TTP) by intent-to-treat analysis was 1.7 months (range, 0.2–21.4 months) for the entire cohort of patients with a PS of 2 (n = 68 patients). The TTP for each of the cohorts was 4.6 months (range, 0.4–14.6 months) for the GC arm (n = 13 patients), 1.5 months for the PCb arm (range, 0.5–13.3 months; n = 15 patients), and 1.4 months for the other two arms (PC: range, 0.2–11.7 months; n = 21 patients; DC: range, 0.2–21.4 months; n = 19 patients). There was no statistically significant difference between the four arms.
The overall median survival and 1-year survival for all 68 patients by intent-to-treat analysis was 4.1 months (range, 0.2–31.0 months) and 19.1%, respectively. The median survival for patients in the PC arm (n = 21 patients) was 7.0 months (range, 0.5–29 months), and the median survival for patients in the GC arm was 7.9 months (range, 0.4–16.2 months; n = 13 patients). The median survival was 2.3 months (range, 0.2–31 months) and 4.6 months (range, 0.9–13.3 months) for patients in the DC arm (n = 19 patients) and the PCb arm (n = 15 patients), respectively (Fig. 1). The 1-year overall survival rates for patients in the PC, GC, DC, and PCb arms were 19%, 38.5%, 10.5%, and 13.3%, respectively. When the log-rank test was applied comparing the overall median survival of each arm, the only statistically significant difference was a survival advantage for patients in the PC arm over patients in the DC arm (P = 0.036).
The 1990s were remarkable for the advent of new agents that have improved survival and response rates compared with cisplatin alone. However, the improvements have been modest. Response rates have improved to about 25–30% in randomized trials, and the median survival has improved from 4 months to 5 months with best supportive care, to 7–8 months with older cisplatin-based combinations, and now hovers between 8 months and 10 months for the newer combinations. The 1-year survival data also have increased from approximately 10% for best supportive care, to 20–25% for the older cisplatin combinations, to about 35% with the newer combinations. These benefits, however, predominantly are restricted to patients with an ECOG PS of 0–1.6, 13–18
This study confirms previous findings that patients with a PS of 2 have a substantial incidence of Grade 3 and 4 hematologic and nonhematologic toxicities. The shorter overall survival of patients who had a PS of 2 compared with patients who had a PS of 0–1 also is consistent with previous observations. There were five deaths (7.35%) noted in this group of patients with a PS of 2, although, on close review, only two of the deaths were attributable directly to therapy. One of the treatment-related deaths may have been due to gemcitabine-induced hemolytic uremic syndrome with renal failure, and the other death was secondary to neutropenic sepsis. The three nontreatment-related deaths were related to lethal consequences of advanced malignancies, namely, brain metastases, thromboembolic events, and exsanguination from hemoptysis. This strongly suggests that the untoward incidence of adverse events that resulted in the termination of this study for patients with a PS of 2 was not all drug-related and was secondary, at least in part, to the comorbidities that often are associated with an impaired PS. Moreover, the toxicity data from the patients who had a PS of 0 or 1 in ECOG Study E1594 did not differ from the cohort of patients who had a PS of 2. It also is noteworthy that the toxicity profile in the PC arm of E5592 that limited enrollment to patients with a PS if 0 or 1 also was not appreciably different. The toxicities in E5592 included a 74% incidence of Grade 4 neutropenia, a 20% incidence of Grade ≥ 3 anemia, a 20% incidence of Grade ≥ 3 neuropathy, and a 4.4% incidence of treatment-related deaths.6
It remains uncertain whether the benefits of combination chemotherapy will extend to patients with a PS of 2. Because of the small numbers of patients in this data set, no definitive statements regarding which of the regimens is preferable can be made. However, it is apparent that cisplatin-based regimens are associated with more nausea and emesis compared with the combination of carboplatin and paclitaxel. By contrast, the trend toward a worse survival with a carboplatin plus paclitaxel regimen warrants further investigation. In other words, despite the diminished toxicity with the combination of carboplatin and paclitaxel, this regimen cannot be recommended as standard treatment for patients with a PS of 2. In fact, no emphatic recommendations can be made for the treatment of these patients off protocol, and combination chemotherapy cannot be considered standard treatment. However, these patients make up a unique group for which clinical trials should be designed to answer specific questions. These include the following: Does chemotherapy confer a survival advantage or provide better palliation compared with best supportive care? Are there well-tolerated regimens in this group? What is the role of targeted biologic therapies? Is combination therapy better than single-agent therapy? What is the influence of specific comorbidities?
The substantial numbers of adverse events that were due only in part to therapy and the overall poor survival in this subset of patients dictate the exploration of alternative approaches. For example, it may be possible to develop less toxic drug combinations. Some investigators already have begun exploring nonplatinum-containing regimens in elderly or medically compromised patients with encouraging results.19–21 Multiple Phase II trials have shown that nonplatinum-containing doublets are active and well tolerated,20, 22, 23 whereas other studies have shown that single agents, such as vinorelbine, gemcitabine, and irinotecan, are as effective as cisplatin alone or older cisplatin-based regimens.14, 15, 17 A study of single-agent vinorelbine compared with best supportive care was conducted in patients age > 70 years.21 In that group, which included approximately 25% of patients with a PS of 2, there was an improvement in palliation, time to progression, overall survival, and quality of life for the group of patients who received vinorelbine. Single-agent therapy may be an alternative strategy. A nonplatinum doublet, gemcitabine plus vinorelbine, recently was compared with vinorelbine alone in patients age > 70 years. Approximately 25% of that population had a PS of 2. There was an increase in the median survival from 4.5 months with vinorelbine alone to 7.0 months with the combination. There was no difference in toxicity, and there was an improvement in quality of life in the patients who were treated with the two-drug combination.24 In contrast, Gridelli et al.25 reported the results of a much larger Phase III study comparing gemcitabine plus vinorelbine versus monotherapy with gemcitabine or vinorelbine in elderly patients. Patients with a PS of 2 comprised 20% of that patient population. Seven hundred patients were randomized, and the response rate was 16.6% for gemcitabine, 18.5% for vinorelbine, and 20% for the combination. The median survival was 28 weeks, 37 weeks, and 32 weeks, respectively, for patients in the gemcitabine, vinorelbine, and combination arms, with respective 1-year survival rates of 26%, 41%, and 31%. The conflicting results from these trials emphasize the perils of interpreting data from unconfirmed Phase III trials. However, the data as a whole clearly demonstrate that modern, single-agent (nonplatinum) chemotherapeutic agents are active and well tolerated in the elderly population and can be evaluated in clinical trials specifically designed for patients with a PS of 2.
Given the nonsignificant trend toward increased median survival in patients in two of the cisplatin doublet arms in E1594 and the low toxicity in patients in the PCb arm, ECOG has embarked on another approach. The current ECOG trial in patients with a PS of 2 is evaluating the use of attenuated doses of two of the current regimens to determine whether the therapeutic window can be widened by reducing the inherent toxicities of therapy. The trial is a randomized Phase II study of carboplatin (AUC of 6) and paclitaxel 200 mg/m2 over 3 hours every 3 weeks and gemcitabine 1000 mg/m2 on Days 1 and 8 with cisplatin 60 mg/m2 on Day 1 of every 3 weeks. These two regimens were chosen for specific reasons: The PCb regimen had the lowest incidence of Grade 3 and 4 toxicities among patients with a PS of 2 in E1594, and the GC combination yielded the highest response rate and a median survival of 7.9 months. It is recognized that this latter finding may be attributed to the randomization of subtle factors, such as proportionally more females and fewer patients with ≥ 10% loss of weight, to this treatment arm.
The doses in the PCb arm were altered. The dose of paclitaxel was decreased from 225 mg/m2 over 3 hours to 200 mg/m2 over 3 hours to minimize the incidence of cumulative neurotoxicity. Support for the use of lower doses of paclitaxel is derived from studies showing that there was no difference in survival between patients who received 175 mg/m2 or 225 mg/m2 26 and that doses > 200 mg/m2 resulted in a higher incidence of clinically significant neuropathy.27–30 Studies using lower doses of gemcitabine and cisplatin also have shown similar activity and less toxicity compared with higher doses.31, 32 In particular, Rosell et al.31 employed 1250 mg/m2 of gemcitabine given 2 of every 3 weeks with 100 mg/m2 of cisplatin on Day 1, and Rinaldi et al.32 administered 1000 mg/m2 of gemcitabine on Days 1 and 8 with cisplatin dosed at either 100 mg/m2 or 70 mg/m2 on Day 2 of a 21-day cycle.
In summary, it was noted in ECOG Study E1594 that the treatment regimens employed were associated with a substantial number of adverse events and a poor overall survival in patients with NSCLC who had a PS of 2. It also should be noted that three of the five deaths in this patient subset were not related to therapy, highlighting the fact that patients with a PS of 2 experience significant comorbidities. These observations support the notion that this group of patients needs special consideration when receiving treatment with chemotherapy. Therefore, ECOG is conducting a randomized Phase II trial in patients with a PS of 2 specifically to evaluate the efficacy and tolerability of gemcitabine plus cisplatin and carboplatin plus paclitaxel combinations with judiciously attenuated doses.