Benefits of Adding a Drug to a Single-Agent or a 2-Agent Chemotherapy Regimen in Advanced Non–Small-Cell Lung Cancer: Title and subTitle BreakA Meta-analysis FREE

Lung cancer is a major cause of mortality worldwide, with an estimated annual incidence of more than 1.2 million cases and mortality of more than 1.1 million cases.¹ Non–small-cell lung cancer (NSCLC) accounts for at least 80% of all lung tumors and approximately two thirds of patients initially present with inoperable NSCLC.^{2 - 3}

During the last 2 decades, numerous randomized clinical trials have investigated the effect of chemotherapy in patients with advanced NSCLC. A meta-analysis based on individual patient data demonstrated that cisplatin-based chemotherapy led to an absolute increase in 1-year survival of 10% in patients with metastatic disease and increased median survival by 2 months.⁴ Cisplatin-based chemotherapy was therefore considered the standard treatment in advanced NSCLC.^{4 - 5} Other meta-analyses on published data showed that systemic chemotherapy prolonged survival moderately compared with the best supportive care.^{6 - 8} Nevertheless, some trials demonstrated that systemic chemotherapy improved quality of life. Although chemotherapy is recommended in patients with good performance status, the question as to whether adding a third agent (triplet) to a 2-drug regimen (doublet), or even a second agent to a single-agent regimen is really beneficial continues to fuel debate.^{9 - 12} A meta-analysis demonstrated that combined chemotherapy increased the objective tumor response rate but led to greater toxicity than single-agent therapy.¹³ The benefit of multidrug regimens was found to be even more controversial in terms of survival.^{3 ,14} The same meta-analysis showed that the benefit of combined chemotherapy in terms of survival was no longer significant when the single agents were platin compounds or vinorelbine.¹³

During the early 1990s, third generation cytotoxic agents yielded promising activity in the treatment of NSCLC in terms of tumor response and tolerance.^{15 - 16} However, a recent review of 22 years of randomized clinical trials on advanced NSCLC in the United States showed that the improvement in median survival between investigational and control arms rarely exceeded 2 months.¹⁴ There is no clear consensus concerning the number and the type of drugs likely to yield better results in terms of tolerance and efficacy.

The aim of the present literature-based meta-analysis was to evaluate the benefit of adding one drug to a single-agent or to a 2-agent chemotherapy regimen (ie, doublet regimen vs single-agent regimen or triplet regimen vs doublet regimen) in terms of the tumor response rate, survival, and toxicity in patients with advanced NSCLC.

Data from all published randomized trials comparing a doublet regimen with a single-agent regimen or comparing a triplet regimen with a doublet regimen in patients with advanced NSCLC were sought electronically. The search period included trials published between January 1980 and October 2003. The computerized bibliographic searches of MEDLINE and EMBASE were performed using the search terms: non-small cell lung carcinoma/drug therapy, adenocarcinoma, large-cell carcinoma, squamous-cell carcinoma, lung neoplasms, clinical trial phase III, and randomized trial. The search results were supplemented with references in books, reviews, journals, and meeting proceedings that were found through hand searches. In addition, we also reviewed the conference proceedings published between January 1982 and October 2003 of the American Society of Clinical Oncology, the European Society of Medical Oncology, the European Cancer Conference, and the International Association for the Study of Lung Cancer. The search was performed without any language restrictions. When an abstract from a meeting and a full article referred to the same trial, only the full article was included in the analysis. When 2 articles or more used the same data, the most updated article was used.

Trials were eligible if they included patients with advanced inoperable NSCLC and compared the benefit of adding a drug to a single-agent regimen or to a doublet regimen. Accrual had to have taken place between January 1, 1980, and December 31, 2001, with unbiased randomization. The methods have been fully described in the protocol available on the Cochrane Library (see http://www.cochrane.org/reviews).

Trials were excluded if they included patients who were candidates for curative surgery, radical radiotherapy, or who had received prior chemotherapy, or included regimens using cisplatin and carboplatin in the same arm, used drugs not considered established cytotoxics, or allowed crossing over between the treatment arms.

Two independent investigators reviewed the publications. When individuals failed to concur, a third investigator reviewed the publication independently. Each trial publication was searched for year of publication, inclusion period, number of patients (randomized and analyzed), disease stage, age, performance status, histological characteristics, objective tumor response rate, median survival, 1-year survival, and specific (hematologic and nonhematologic) toxicity. When available, we recorded information on the randomization process and whether the trial used an intent-to-treat (ITT) analysis. The primary end point was the overall tumor response rate. The secondary end points were overall survival, whose evaluation was based on the 1-year survival rate and median survival and toxicity.

Some trials compared both a doublet regimen with a single-agent regimen and a triplet regimen with a doublet regimen. When a trial compared more than 2 different chemotherapy regimens, the control or the investigational arm was counted twice or more in the analysis. Consequently, more comparisons than trials were studied. Third-generation cytotoxic agents termed newer drugs include vinorelbine, gemcitabine, docetaxel, paclitaxel, and tirapazamine. Newer drugs were analyzed separately from first- or second-generation drugs termed older drugs. Cisplatin is considered more efficient than other older drugs, such as carboplatin. The activity of carboplatin is not recognized worldwide as equivalent to that of cisplatin.^{17 - 18} However, because of the small number of trials with other older drugs, we decided to keep all older drugs together. According to the type of chemotherapy in the control arm, 2 categories of trials were identified: older drug-based and newer drug-based regimens. According to the type of chemotherapy added in the investigational arm, the trial categories were addition of an older drug and a newer drug. When the control arm was a doublet, regimens with at least 1 newer drug were considered newer drug-based regimens.

All analyses were performed using ITT when possible. When the number of randomized patients in each arm was available, it was used as a denominator to compute the tumor response and toxicity rates. For survival, we used the number reported by the investigators. When the total number of randomized patients was different from the sum of the numbers of patients analyzed per arm, an ITT analysis was noted as nonapplicable but the trial was still included in the calculations of the meta-analysis. When the number of patients per arm was not available and the results were expressed as percentages, the trials were excluded.

Treatment effects were displayed through odds ratios (ORs) for tumor response, 1-year survival, and toxicity and through median ratios (MRs) for median survival. For 1-year survival, the OR for death of doublet regimen over single-agent regimen (triplet regimen over doublet regimen) was used. The control arm was used as reference category for the computation of ratios, except for tumor response for which it was the experimental arm. Consequently, less than unity (<1.0) for ORs and MRs indicates benefit of the doublet regimen over the single-agent regimen (triplet regimen over doublet regimen) and corresponds respectively to a higher tumor response rate, a higher survival rate, less toxicity, and a prolonged median survival. The ORs for the treatment effect were calculated for individual trials. For each comparison (doublet regimen vs single-agent regimen; triplet regimen vs doublet regimen), a pooled OR was calculated using a fixed-effect model for the tumor response rate, 1-year survival rate, and toxicity rates. The Mantel-Haenszel test stratified by trial was used for the treatment effect and the Cochran test was used for heterogeneity.¹⁹ The ORs for tumor response and survival for the doublet regimen vs the single-agent regimen and for the triplet regimen vs the doublet regimen were compared using an interaction test. An additional analysis using median survival was also used for survival.²⁰ To facilitate interpretations, an absolute benefit between the control and the investigational arms was computed using the overall OR and the rate in the control arm.²¹

As trial designs were expected to be heterogeneous, several factors were studied to explore any significant heterogeneity in outcomes: type of regimen in the control arm; type of drugs added; planned dose reduction in the investigational arm of the drugs common to both arms; and the use of ITT analysis. The test of heterogeneity between groups of trials defined by these factors was called test of interaction. Residual heterogeneity was tested using the difference between the Cochran test for global heterogeneity and the test for interaction. All P values are 2-sided.

One hundred and two randomized trials including 21 051 patients with advanced NSCLC were initially identified, from which 37 trials were secondarily considered ineligible for different reasons (Figure 1). In total, 65 trials (13 601 patients) were considered eligible. Eight trials were excluded because the number of patients per arm was not available. Among them, 6 were reported as abstracts. A full description of the ineligible and excluded trials or trial arms is available from the authors on request.

After the selection procedure (Figure 1), 57 trials including 11 160 patients were analyzed. Twenty of the analyzed trials were reported as abstracts. As several trials had more than 2 eligible arms, the number of comparisons was 68 (11 989 patients). The analyzed comparisons of doublet regimens with single-agent regimens are described in Table 1 and Table 2. The control regimen was cisplatin in 11 comparisons, carboplatin in one, another older drug in 8, and a newer drug in 13. The drug added was cisplatin in 14 comparisons, carboplatin in 2, another older drug in 9, and a newer drug in 8. The analyzed comparisons of triplet regimens with doublet regimens are described in Table 3 and Table 4. The control regimen included cisplatin in 25 comparisons, another older drug in 1 comparison, and at least 1 newer drug in 9. The drug added was cisplatin in 2, carboplatin in 1 comparison, another older drug in 27, and a newer drug in 5.

Data on the tumor response rate was available for all patients, median survival for 88% of patients, and 1-year survival rates were available for 53% of the patients in the analyzed comparisons. The availability of toxicity data was variable; neutropenia toxicity was the information most frequently reported in 61% of the patients.

Objective Tumor Response Rate. Concerning the tumor response rate, 33 (7175 patients) comparisons were available from 29 analyzed trials. Adding a drug to a single-agent regimen was associated with a statistically significant increase in the objective tumor response rate (OR, 0.42; 95% confidence interval [CI], 0.37-0.47; P<.001; Table 5). The absolute benefit was 13%, which corresponds to a 2-fold increase in the tumor response rate from 13% with a single-agent regimen to 26% with a doublet regimen (Figure 2). The benefit was higher when the control arm was an older drug (OR, 0.35) than when it was a newer drug (OR, 0.52) (P = .001; Figure 2). The benefit was also higher when the added drug was older (OR, 0.38) than when it was newer (OR, 0.50), but not significantly (P = .06). Heterogeneity of the treatment effect was found between the comparisons (P<.001; Figure 2). Two trials had a low tumor response rate in the control arm, probably due to the choice of the drug (vindesine for Luedke et al³⁴ and teniposide for Splinter et al²⁵ ). One trial by Gil Deza et al²⁹ had an unusual high tumor response rate with vinorelbine alone. After exclusion of these trials, heterogeneity was no longer significant (P = .17) and the overall OR was similar (0.43; P<.001). The benefit was lower in the 7 comparisons with dose reduction in the experimental arms (OR, 0.62) than in those without dose reduction (OR, 0.36; P<.001). No significant difference was noted between comparisons using an ITT analysis and those that had not (P = .27).

Survival. Thirteen comparisons were available for 1-year survival and 30 for median survival. Adding a drug to a single-agent regimen was associated with a significant increase in 1-year survival (OR, 0.80; 95% CI, 0.70-0.91; P<.001; Figure not shown, available on request). The absolute benefit was 5%, which corresponds to an increase in 1-year survival from 30% with a single-agent regimen to 35% with a doublet regimen (Table 5). Adding a drug to a single-agent regimen significantly increased median survival (MR, 0.83; 95% CI, 0.79-0.89; P<.001, Figure 3). The benefit was higher when the control arm was an older drug than when it was a newer drug for both 1-year survival rate (P = .03) and median survival (P = .007). Heterogeneity of the treatment effect was found between the comparisons for both 1-year survival (P = .002) and median survival (P<.001). The benefit was lower in the comparison with dose reduction in the experimental arms (OR, 1.12; MR, 0.95) than in those without (OR, 0.64; MR, 0.78) for both 1-year survival rate (P<.001) and median survival (P<.001). When the heterogeneity related to dose reduction was taken into account, there was no more residual heterogeneity (P = .64) for 1-year survival. No significant difference was noted between comparisons using an ITT analysis and those trials that had not for 1-year survival (P = .16).

Toxicity. As expected, the rates of grades 3 and 4 toxicity caused by doublet regimens was statistically increased compared with rates following single-agent therapy, with ORs ranging from 1.2 to 6.2 (Table 6). There was no increase in infection rates in doublet regimens.

Objective Tumor Response Rate. Concerning the tumor response rate, 35 comparisons (4814 patients) were analyzed from 28 trials. Adding a third drug to a doublet regimen was associated with a significantly increased objective tumor response rate (OR, 0.66; 95% CI, 0.58-0.75; P<.001; Figure 4). The absolute benefit was 8%, which corresponds to an increase in tumor response from 23% with a doublet regimen to 31% with a triplet regimen (Table 5). There was no difference in tumor response rate whether the doublet regimens contained older or newer drugs (P = .33). There was no significant difference (P = .70) in terms of benefit according to the drug added (older drug OR of 0.64 vs newer drug OR of 0.74). The heterogeneity of the treatment effect between the comparisons was borderline (P = .06; Figure 4). In 12 comparisons, investigators planned to decrease the drug doses in the triplet regimen to avoid additional toxicity. A significant difference was demonstrated between comparisons with (OR, 0.76) and without (OR, 0.57) a dose reduction (P = .03). No significant difference was noted between comparisons using an ITT analysis and those trials that had not (P = .40).

Survival. Ten comparisons, all using ITT analysis, were available for 1-year survival. Data for 30 comparisons were available for median survival. Adding a third drug to a doublet regimen did not yield any additional benefit for 1-year survival (OR, 1.01; 95% CI, 0.85-1.21; P = .88; Figure not shown, available on request) or for median survival (MR, 1.00; 95% CI, 0.94-1.06; P = .97; Figure 5). There was no significant difference according to the type of control regimens used (older drugs vs newer drugs) for both 1-year survival rate (P = .28) and median survival (P = .36). The data did not show heterogeneity for 1-year survival (P = .59), but did show borderline heterogeneity for median survival (P = .04). No difference was observed between comparisons with or without dose reduction for 1-year survival (P = .30) and for median survival (P = .16).

Toxicity. Toxicity of grades 3 and 4 occurred more frequently in triplet regimens than in doublet regimens with ORs ranging from 1.4 to 2.9, except for neurological, renal, auditory, and gastrointestinal toxic effects (Table 6).

A logical strategy for improving the treatment of inoperable NSCLC with cytotoxics is to combine agents whose mechanism of action and toxicity differ and do not overlap. However, whether more chemotherapy is better and how much of it is best for patients with advanced NSCLC is still debatable. The rationale for using multidrug combinations is based on preclinical studies that demonstrated synergism between different cytotoxics in vitro. Synergistic and additive effects have indeed been observed with different combinations.^{80 - 81} This led to the development of cytotoxic regimens combining 2 or 3 agents with different mechanisms of action.⁸² We designed the present meta-analysis to ascertain whether more chemotherapy is better for patients with advanced NSCLC.

Combining newer drugs appeared to be an attractive approach because they showed a clear benefit when used as single agents in terms of efficacy and lower toxicity.⁸³ Currently, most cancer practitioners use platinum-based doublet chemotherapy as first-line treatment of patients with stage IIIB and IV NSCLC. The present meta-analysis provides data in favor of such an approach. Indeed, the data in the present study demonstrate that adding a second agent to a single-agent regimen yields a significant benefit in terms of tumor response and survival. The tumor response rate was doubled when a second drug was added to a single-agent regimen and led to a significant increase in survival rate. This compelling and valuable information supports current clinical practice. However, the present meta-analysis did not address major questions facing the medical community. For example, is a doublet regimen better than a single-agent regimen for patients with a poor performance status and/or in elderly patients? In that regard, there is no consensus in the literature; many clinicians opted for a single-agent regimen for a performance score of 2 or for elderly patients, while others believed that a standard approach should be preferred in such patients.^{48 ,84 - 87}

Although less commonly accepted, triplet combinations are used by several teams throughout the world. Some trials suggested a significant increase in the tumor response rate and median survival with triplet regimens and had acceptable levels of toxicity.⁷⁶ Another trial did not confirm these results.⁵⁸ The data from the present meta-analysis do not support the triplet regimen approach. A moderate but significantly increased tumor response rate was noted when a third drug was added to a doublet regimen, but this higher tumor response rate did not translate into a benefit in terms of survival. However, data on 1-year survival were available in 47% of the patients and definite conclusions on survival are difficult to draw. The OR of tumor responses and 1-year survival rates for doublet regimens compared with single-agent regimens were significantly different (P<.001 and P<.04, respectively) from those for triplet regimens compared with doublet regimens (Table 5). In this study, there was a correlation between the tumor response rate in the control arm and the OR for tumor response (data not shown). The benefit was lower in trials with a high tumor response rate in the control arm than in those with a low tumor response rate. This relationship may explain why the addition of a second drug led to better results than the addition of a third drug. Trials with a planned decreased dose of chemotherapy in the experimental arms had a significantly lower benefit in terms of tumor response rate for both comparisons (doublet regimen vs single-agent regimen and triplet regimen vs doublet regimen) and for survival for a doublet regimen compared with a single-agent regimen. No difference was noted with and without dose reduction for survival for triplet regimens compared with doublet regimens.

As expected, toxicity was significantly more severe in patients who received combination chemotherapy. Symptomatic improvement due to tumor shrinkage should be balanced with increased toxic effects. Unfortunately, data on quality of life, which might be used to balance the clinical benefit of chemotherapy, were rarely available in these trials and no conclusions could be drawn. However, the increased toxicity was relatively lower when a third drug was added to a doublet regimen compared with when a second drug was added to a single-agent regimen. A possible explanation was that when a third drug was added, decreased doses of the drugs common to both arms had been planned in the investigational arm in 12 (41% of the patients) of 35 comparisons compared with only 7 (27% of the patients) of 30 comparisons when a second drug was added to single-agent chemotherapy regimen. The increase in toxicity was less important in the trials with dose reduction than in those without dose reduction for both a doublet regimen compared with a single-agent regimen and a triplet regimen compared with a doublet regimen (data not shown). The imbalance in comparisons with dose reduction may explain the observed difference in survival between the addition of a second drug and of a third drug.

Whether the lack of improvement in survival when a third drug was added to a doublet regimen could be explained by different rates of patients treated with second-line chemotherapy remains unknown. Because triplet regimens yielded better tumor response rates, the clinical value of such results deserves debate. Does a better tumor response rate translate into better quality of life? If such is the case, then a triplet regimen approach could be considered beneficial in patients with stage IV cancer. If not, it might be more reasonable to capitalize on the higher tumor response rates yielded by triplet regimens by treating patients with earlier stage disease (eg, neoadjuvant therapy).

Some trials demonstrated that the survival benefit obtained with chemotherapy appeared to translate into an improvement in the quality of life.⁸⁸ If triplet regimens are to be developed, the mechanism of action of the third drug should be different as well as the targeted cells to achieve synergism. Unfortunately, to date, this has not been demonstrated with the newer targeted agents. Adding the epidermal growth factor inhibitor (gefitinib) to standard chemotherapy failed to demonstrate synergy in triplet regimens compared with a standard doublet regimen in 2 large phase 3 studies.^{89 - 90} The matrix metalloproteinase inhibitor and trastuzumab also failed to improve results.^{91 - 92}

The data reported herein can be contrasted with the rest of the literature. Based on 28 trials and using a different methodological approach, Baggstrom et al⁹³ also concluded that tumor response and survival were improved by adding a second drug to a single-agent regimen. They were unable to show any advantage in adding a third drug to a doublet regimen. Their study was limited to newer drugs and to trials published in English. Differences in risk were used instead of ORs. Although these meta-analyses used different statistical methods, their results are consistent. A previous literature-based meta-analysis by the same authors suggested that newer drugs resulted in better overall tumor response and 1-year survival compared with older drugs.⁹⁴ Our results showed that when the single-agent regimen consisted of a newer drug, the benefit gained in adding a second agent was significantly lower than when the single agent was an older drug. This suggests that newer agents were also more efficient in the single-agent setting. When triplet regimens were compared with doublet regimens and when the control arm contained a newer or an older agent, adding a third agent yielded a similar benefit in terms of tumor response. Most of the older doublet regimens were based on the use of cisplatin. This supports the hypothesis that cisplatin or third generation–based doublet regimens might be equally efficient. However, no definite conclusions can be drawn from our data on the superiority of third generation agents and further trials are needed to explore this question.

The present study had limitations. A meta-analysis based on published data might overestimate the treatment effect because such a method does not allow researchers to carefully double check data and conduct detailed survival analyses.⁹⁵ The main limitation of our analysis was the amount of missing data. Data on median survival were reported in only 88% of the patients and 1-year survival in 53% of the patients.

Significant heterogeneity was observed in the treatment effect between the trials. Drug efficacy improved during the period in which the trials were conducted (between 1980 and 2001), mainly due to the emergence of newer agents.^{15 - 16} We showed that part of the heterogeneity might be due to dose reduction in some trials. The heterogeneity was no longer significant in a doublet regimen compared with a single-agent regimen for 1-year survival when the heterogeneity related to dose reduction was taken into account. Finally, we used median survival to analyze survival because this was reported more frequently in the analyzed trials. This method might lead to biased results compared with hazard ratios based on individual patient data, which estimate the overall treatment effect on survival.⁹⁶

In conclusion, our data do not provide any rationale for the use of triplet regimens in inoperable NSCLC and their use should be strictly limited to clinical trials.