Second-Line Treatments in Non-small Cell Lung Cancer: A Systematic Review of Literature and Metaanalysis of Randomized Clinical Trials FREE TO VIEW

In the 10 years, the treatment of metastatic non-small cell lung cancer (NSCLC) in progression after first-line chemotherapy has changed, and docetaxel therapy every 3 weeks has been identified as the “gold standard” for patients with good performance status.^1–7 However, further trials comparing docetaxel therapy every 3 weeks with weekly docetaxel therapy^8–11 (superiority trials) or pemetrexed¹² and oral topotecan¹³ (noninferiority trials) have shown equivalent efficacy (superiority trials) or “noninferior” efficacy (noninferiority trials) with a better safety profile of the novel schedules in comparison with docetaxel therapy every 3 weeks. Moreover, the clinical scenario has recently been enlarged by the use of a novel class of molecules that act as epithelial growth factor receptor (EGFR) inhibitors (in particular gefitinib and erlotinib).^14–17 However, despite the evidence supporting the use of second-line treatments in all patients who progress after a platinum-containing regimen, the modest advantage in survival (a few months in comparison with placebo), the negligible response rate (RR) [usually < 10%], the relevant side effects, and the nonunivocal data about symptoms control or quality of life improvement seem to reduce their role in clinical practice.^14,15

In our systematic review of literature with metaanalysis of randomized clinical trials, we tried to define the role of second-line treatments in advanced NSCLC, identifying two main questions to be solved: (1) to define the role of second-line treatments in advanced NSCLC in comparison with best supportive care (BSC); and (2) to assess the role of alternative treatments to docetaxel therapy every 3 weeks, assuming docetaxel therapy every 3 weeks as the standard second-line treatment of advanced NSCLC.

A systematic review of MEDLINE, EMBASE, CINAHL, CRISP, and Cochrane Systematic Reviews databases from January 2000 to July 2008 was independently performed by two authors (D.T. and L.L.A.), using “Carcinoma, Non Small-Cell Lung” (medical subject heading [MeSH] term) and “Drug Therapy” (MeSH term) and “docetaxel” (MeSH term) or “erlotinib” (MeSH term) or “gefitinib” (MeSH term) as search terms. A primary analysis was planned to compare the outcomes of antineoplastic treatments vs BSC, and a secondary analysis was performed to compare any alternative antineoplastic treatment with three times per week docetaxel, assumed as the “gold standard.” No language restriction in trial searching was applied. Likewise, a systematic review of the Central Registry of Controlled Trials of the Cochrane Library and the National Cancer Institute Database of Clinical Trials was independently performed by the same authors with the last search update in July 2008. One-year survival rate (SR) in the primary analysis was the primary end point; RR and time to progression in the primary analysis, and 1-year SR, RR, and time to progression in the secondary analysis were the secondary end points. Neither symptoms control nor quality-of-life improvement was included in the analysis because of the intrinsic difficulties in making the data about symptoms control or quality of life comparable in a pooled analysis. Likewise, no data about side effects or tolerability of the treatments were extracted from the trials included in the primary or secondary analysis, and therefore no pooled analysis of side effects was performed. All randomized phase III trials comparing any antineoplastic treatment with BSC were considered eligible for the primary analysis; all randomized phase III trials comparing two or more antineoplastic treatments were considered eligible for the secondary analysis. All the references of the retrieved trials and all the systematic reviews published from January 2000 to July 2008 were analyzed, and all retrieved trials for the primary or secondary analysis were included if eligible. When discrepancies in trials searching occurred between the two independent reviewers (D.T. and L.L.A.), they were discussed with a third researcher (E.T.) to reach a final consensus. All the selected trials published as full-text articles in a peer-reviewed journal were analyzed and classified using the Nicolucci score,¹⁸ whereas the data of congress proceedings not published as full text articles were included in the analysis if appropriate, but were not classified with the Nicolucci score. Table 1 details the inclusion and exclusion criteria. For all the selected trials, the following items were recorded (when available): the first author, year of publication, sample size and patient distribution in the arms of treatment, patients age, sex, “smoker status” response to first-line chemotherapy, Eastern Cooperative Oncology Group (ECOG) performance status at the time of entering the trial, histologic characterization of the disease, design of the trial and methods of randomization (if blinded and centralized), kind of second-line treatment, assessment of primary and secondary end points, and number and reasons for missing data. All these data were used to stratify the results in the metaregression analysis and to define the quality of the trials with the Nicolucci score. The qualitative and quantitative analysis of the selected articles was independently performed by two authors (D.T. and E.T.). When discrepancies occurred, they were discussed with a third researcher (L.L.A.) to reach a final consensus.

An analysis of patient distribution in the selected trials based on age, sex, histologic type, ECOG performance status, and best previous response to first-line chemotherapy was performed with a descriptive aim when possible. Heterogeneity among the trials was assessed with descriptive aim using the I² test; I² values > 50% were deemed to suggest large among-trial heterogeneity, values of 25 to 50% were deemed to show modest heterogeneity, and values < 25% were deemed to represent low heterogeneity. When possible, methodological heterogeneity and clinical heterogeneity were separately investigated either comparing the trial design in the primary and secondary analysis or using the metaregression model with mixed effects (when applicable) for age, sex, ECOG performance status, and RR to first-line treatments. An α level < 5% was assumed to be statistically significant.

The pooled analysis of the odd ratios (ORs) was performed using a random effect model, and an α level < 5% was assumed statistically significant. A subanalysis, including only the data extracted from the full-text articles published in peer-reviewed journals and classified by the Nicolucci score, was also performed with descriptive aim. Publication bias was investigated in the primary and secondary analysis with the qualitative instrument of the funnel plot and the quantitative Egger regression intercept model. An α level < 5% was assumed statistically significant.

We retrieved and examined 336 descriptive reviews, 104 clinical trials, and 8 systematic reviews with metaanalysis in the preliminary evaluation. Fifteen of these 448 articles (all of them reporting data of randomized phase III clinical trials) were considered potentially eligible because they reported mature data comparing different therapeutic options in second-line treatment of NSCLC, and 14 were included into the analysis. Two of the 15 potentially eligible studies^3,19 reported data from the same trial (the TAX317 trial) and were analyzed as a single report. Three trials comparing antineoplastic treatments with BSC were included in the primary analysis,^3,16,17 and 11 trials^{7–13,20–24} comparing docetaxel administration every 3 weeks with other docetaxel schedules or ifosfamide/vinorelbine, pemetrexed, oral topotecan, paclitaxel poliglumex, vinflunine, gefitinib, or docetaxel-gemcitabine combination were included in the secondary analysis. One trial published in two different journals^3,19 and included in the primary analysis as a single report (TAX317 trial) compared docetaxel administration every 3 weeks with BSC, one trial¹⁶ compared gefitinib with placebo (the Iressa Survival Evalaution in Lung Cancer [ISEL] trial), and one trial¹⁷ compared erlotinib with placebo (BR-21 trial). As concerns the secondary analysis, docetaxel every 3 weeks was compared with ifosfamide or vinorelbine in one trial (TAX320 trial),⁷ with weekly docetaxel in three trials,^8–10 with pemetrexed in one trial,¹² with oral topotecan in one trial,¹³ with paclitaxel poliglumex in one trial,²⁰ with vinflunine in one trial,²¹ with gefitinib in two trials,^22,23 and with the docetaxel-gemcitabine combination in one trial.²⁴ Six trials included in the secondary analysis were published as full-text articles in peer-reviewed journals,^7–10,12,13 and five trials were published as abstracts in congress proceedings.^20–24 The main characteristics of the trials and main inclusion criteria are detailed in Table 2. The Nicolucci score of the nine eligible full-text published trials ranged from 54 to 89%, with a median value of 70% (Table 2). The recruitment in the trials included in the primary analysis began in 1994 (TAX317 trial)³ and concluded in 2004 (ISEL trial)¹⁶; the recruitment in the trials included in the secondary analysis began in 1995 (TAX320 trial)⁷ and concluded in 2003 (the trials of Schuette et al,⁹ Camps et al,¹⁰ and Ramlau et al¹³). All the trials published as full-text articles reported data about age, sex, and ECOG performance status of the patients.^{3,7–13,16,17} The RR to first-line chemotherapy was reported in the TAX317,³ ISEL,¹⁶ BR21,¹⁷ Gridelli et al,⁸ and Hanna et al¹² trials; in the TAX320 trial,⁷ it was reported as a complete and partial response plus stable disease. The histologic type was reported in the ISEL,¹⁶ BR21,¹⁷ Schuette et al,⁹ Chen et al,¹¹ Hanna et al,¹² and Ramlau et al¹³ trials.^{9,11–13,16,17} The “never-smoker” status was reported in the ISEL¹⁶ and BR21 trials.¹⁷ The main prognostic factors reported in the trials published as full-text articles are detailed in Table 3. No univocal data about the main prognostic factor could be extracted from congress proceedings. The outcomes of 2,627 patients were analyzed in the primary analysis, and those of 5,952 patients in the secondary analysis (Fig 1).

Three trials for a total of 2,627 patients were included in the primary analysis. The TAX317 trial (204 patients) compared docetaxel plus BSC vs BSC alone, the ISEL trial (1,692 patients) compared gefitinib plus BSC vs BSC alone, and the BR21 trial (731 patients) compared erlotinib plus BSC vs BSC. The TAX317 trial included patients treated with either docetaxel, 100 mg/m² every 21 days, or docetaxel, 75 mg/m² every 21 days, according to the changes in the protocol design during the conduction of the trial. One-year SR and RR were reported in all the trials, and time to progression in the BR21 and ISEL trials. Significant heterogeneity among the trials was documented for 1-year SR either including all patients or excluding the patients treated with docetaxel, 100 mg/m² every 21 days, in the TAX317 trial (I² = 60.157%, p = 0.081, and I² = 73.15%, p = 0.024, respectively). The sole methodological difference among the three trials was the use of chemotherapy in the TAX317 trial and an EGFR inhibitor in the ISEL and BR21 trials; no further potential causes of methodological heterogeneity were found in the design and conduction of the trials. A significant influence in clinical heterogeneity among the trials was identified with the metaregression model for age (R = 0.125; 95% confidence interval [CI], 0.006 to 0.243; p = 0.038), sex (R = 0.125; 95% CI, 0.006 to 0.243; p = 0.038), and RR to first-line chemotherapy (R = − 0.011; 95% CI, − 0.022 to − 0.001; p = 0.02). A trend toward the statistical significance was observed for ECOG performance status (R = − 0.07; 95% CI, − 0.147 to 0.007; p = 0.075). The RR to first-line chemotherapy of the ISEL trial was significantly lower than that of the TAX317 and BR21 trials (18.5%, 34%, and 40%, respectively) [Table 3]. The odds ratio (OR) of 1-year SR was 0.773 (p = 0.021) and 0.736 (p = 0.029), respectively, with an absolute improvement in 1-year SR of 7% (Fig 2,, top, A). No heterogeneity among the trials was documented for RR either including all patients or excluding those treated with docetaxel, 100 mg/m² (I² = 0%; p = 0.48). The OR of RR was 0.165 (p < 0.001) and 0.166 (p < 0.001), respectively (Fig 2, bottom, B). The publication bias of the primary analysis (α-intercept = − 2.89; 95% CI, − 14.00 to 8.21; p = 0.34) is shown in Figure 3, top, A.

Significant heterogeneity between the two trials investigating the EGFR inhibitors^16,17 was documented for 1-year SR (I² = 74.59%; p = 0.047), with an OR of 0.797 (p = 0.057). Low heterogeneity between the two trials was documented for RR (I² = 13.66; p = 0.282), with an OR of 0.164 (p < 0.001), and significant heterogeneity was documented for time to progression (I² = 85.718%; p = 0.008), with an OR of 0.714 (p = 0.022) [Fig 4].

Eleven trials,^{7–10,12,13,16,17,20–24} for a total of 5,952 patients, were included in the secondary analysis. There were 694 patients enrolled in the three trials^8–10 comparing docetaxel therapy every 3 weeks with weekly docetaxel therapy, 571 patients in the trial comparing docetaxel therapy every 3 weeks with pemetrexed docetaxel therapy,¹² 829 patients in the trial comparing docetaxel therapy every 3 weeks with oral topotecan therapy,¹³ 849 patients in the trial comparing docetaxel therapy every 3 weeks with paclitaxel poliglumex therapy,²⁰ 551 patients in the trial comparing docetaxel therapy every 3 weeks with vinflunine therapy,²¹ 1,955 patients in the two trials comparing docetaxel therapy every 3 weeks with gefitinib therapy,^22,23 and 130 patients in the trial comparing docetaxel therapy every 3 weeks with the docetaxel-gemcitabine combination therapy.²⁴ The TAX320 trial included patients treated with either docetaxel, 100 mg/m² every 21 days, or docetaxel, 75 mg/m² every 21 days, according to the changes in the protocol design during the conduction of the trial. One-year SR was reported in all trials, RR in all trials but one (the Bonomi et al²⁰ trial), and time to progression only in the Hanna et al¹² and Ramlau et al¹³ trials. Only a cumulative statistical analysis of heterogeneity could be performed including all the selected trials into the analysis, whereas a separate subanalysis of the methodological and clinical heterogeneities could be performed including just the six articles published as full text in peer-reviewed journals. A modest heterogeneity among the trials was documented for 1-year SR either including all patients or excluding those treated with docetaxel, 100 mg/m² every 21 days, in the TAX317 trial (I² = 27.324%; p = 0.184; and I² = 38.273%, p = 0.094, respectively). The overall OR of 1-year SR was 0.924 (p = 0.122) and 0.922 (p = 0.256), respectively. The OR of 1-year SR was 0.957 (p = 0.605), 0.949 (p = 0.584), and 0.877 (p = 0.165) analyzing all patients treated with chemotherapy, all patients treated with chemotherapy except those treated with docetaxel, 100 mg/m² every 3 weeks, and the patients treated with EGFR inhibitors, respectively. Modest heterogeneity among the trials was documented for RR either including all patients or excluding those treated with docetaxel, 100 mg/m² (I² = 42.713%; p = 0.073 and I² = 35.952%, p = 0.120, respectively). The overall OR of RR was 1.048 (p = 0.761) and 1.069 (p = 0.643), respectively. The OR of RR was 0.879 (p = 0.461), 0.902 (p = 0.508), and 1.426 (p = 0.219) analyzing all patients treated with chemotherapy, all the patients treated with chemotherapy except those treated with docetaxel, 100 mg/m² every 3 weeks, and the patients treated with EGFR inhibitors. The publication bias (α-intercept = 0.414; 95% CI, − 1.392 to 2.22; p = 0.684) is shown in Figure 3, bottom, B.

One-year SR and RR were also analyzed in a subanalysis including only the six trials published as full-text articles in peer-reviewed journals (2,628 enrolled patients). Significant heterogeneity among the trials was documented for 1-year SR either including all patients or excluding those treated with docetaxel, 100 mg/m² every 21 days, in the TAX317 trial (I² = 49.305%; p = 0.066; and I² = 58.394%, p = 0.025, respectively). Two differences among the six trials can be identified as factors of methodological heterogeneity:

No further factors of methodological heterogeneity could be found in the design and conduction of the six trials included in the subanalysis. A significant influence in clinical heterogeneity among the trials was identified with the metaregression model for age (R = 0.119, 95% CI, 0.025 to 0.213; p = 0.01), and a trend toward the statistical significance was observed for ECOG performance status (R = 0.1, 95% CI, − 0.017 to 0.218; p = 0.095). No influence was observed for gender (R = 0.01, 95% CI, − 0.025 to 0.048; p = 0.537). RR to first-line chemotherapy was not investigated because of the nonunivocal data reported in the six articles.

The OR of 1-year SR was 1.01 (p = 0.943) and 0.991 (p = 0.954), respectively. Modest heterogeneity among the trials was documented for RR either including all patients or excluding those treated with docetaxel, 100 mg/m² (I² = 39.637%, p = 0.127 and I² = 28.864%, p = 0.20, respectively). The OR of RR was 0.9 (p = 0.661) and 0.923 (p = 0.713), respectively. Significant heterogeneity among the trials was documented for time to progression in the two trials reporting this datum^12,13 (I² = 0.269%, p = 0.067), with an OR of 0.921 (p = 0.437). All results of primary and secondary analysis are detailed in Tables 4 and 5.

In the last few years, many authors have investigated the role of second-line therapies in advanced NSCLC,^{14,15,25–27} and different options were recommended by the primary scientific societies as the “gold standard.”^1,2,4,5 Based on these recommendations, docetaxel or pemetrexed represent the main chemotherapeutic options, and erlotinib is the biological alternative to chemotherapy, mainly in no-smoker females with adenocarcinoma.^{14,15,25–27} However, some issues remain controversial and merit to be analyzed in detail.

The first aspect is the identification of the clinical outcomes of a treatment, to define the efficacy of the different options. Our primary analysis can probably answer this question. The three trials selected for the primary analysis^3,16,17 can be classified as high-quality trials by the Nicolucci score, and all of them (in the pooled analysis) demonstrated a significant improvement in overall survival in comparison with BSC alone. Such a result is innovative because it identifies the potential effect of a treatment (chemotherapy or EGFR inhibitors) in a hard clinical outcome. Interestingly, either the comprehensive analysis including all three trials or the subanalysis including only the EGFR inhibitors confirm that any treatment against the tumor is superior to BSC. It follows that both therapy every 3 weeks with docetaxel or EGFR inhibitors can be considered the “gold standard” in the treatment of these patients. However, although therapy every 3 weeks with docetaxel is the sole chemotherapeutic option showing better overall survival than BSC, erlotinib (but not gefitinib) significantly improved overall survival when the three trials were analyzed separately,^3,16,17 favoring different recommendations for clinical use by the European and American health regulatory agencies.^28–32 Our analysis yields just a pooled result and does not allow a separate comparison of the outcome among the three trials. However, the differences in the main prognostic factors reported in the trials (in particular the lower RR to first-line chemotherapy observed in the ISEL trial) could justify the different outcome of the ISEL trial in comparison with the TAX317 and BR21 trials, as well as the differences observed in our pooled analysis. In our analysis, both the cumulative data of second-line treatments and those of the treatments with EGFR inhibitors seem to demonstrate a significant advantage of the treatments, and to overcome the limits suggested by the final results of the single trials. However, even though significant from a statistical point of view, the absolute 1-year SR improvement for treatments in comparison with BSC is approximately 7%, and the consequent number needed to treat (NNT) for 1-year-patient-alive gained is 14. Such a high NNT does not reduce the statistical and speculative significance of our analysis but limits its clinical significance, and it suggests the need of a preselection of patients to be treated with second-line treatments. A clinical and biological selection of the patients might favor more impressive improvement of the intrinsic efficacy of second-line treatments, as demonstrated by Weiss et al for chemotherapy,³³ and Thatcher et al¹⁶ or Shepherd et al¹⁷ for EGFR inhibitors. If only patients who obtained a significant advantage from first-line chemotherapy or patients with specific clinical or biological characteristics seem to have a significant response to second-line treatments, further studies should be addressed to obtain a validation of these profiles, to allow a selection of potentially responder patients and a reduction of the NNT for the single treatments.^33–49 If such a prospect of preselecting the patients to be treated with second-line treatments is suggestive for chemotherapy, it is even more intriguing for EGFR inhibitors, as suggested by much recent data that demonstrated a greater advantage in the main outcomes for patients with peculiar biological or clinical characteristics in comparison with unselected patients.^34–49

Two further considerations merit attention about the outcomes of our primary analysis:

Essentially, two suggestions can be inferred from our primary analysis: second-line treatments may favor an improvement of overall survival, and therapy every 3 weeks with docetaxel or EGFR inhibitors should be considered the treatment of choice. However, the low improvement of 1-year SR and the high NNT suggest the need of a preselection of potentially responder patients, to improve the outcome and to reduce the NNT of the treatment.

The head-to-head comparison among the different options of treatment reported in our analysis also has some interest because it can yield some useful information for clinical practice. Some phase III randomized trials^7–13 compared therapy every 3 weeks with docetaxel with other chemotherapeutic options, and a metaanalysis⁵⁰ compared three times per week docetaxel with weekly docetaxel. Furthermore, two randomized phase III trials^22,23 and a randomized phase II trial,⁵¹ comparing gefitinib with docetaxel therapy every 3 weeks, dealt with the comparison between chemotherapy and target therapies.

It is known that the TAX317 and the TAX320 trials identified docetaxel therapy every 3 weeks as the “gold standard” in the treatment of platinum-resistant NSCLC but reported a significant toxicity.^3,7 Consequently, all subsequent trials^{7–13,20–24,51} were aimed at identifying new schedules with comparable activity and a better safety profile than docetaxel therapy every 3 weeks. Our analysis, focused on the outcomes and not on the safety of second-line treatments, confirms the comparable efficacy of all the schedules but does not identify any schedule as preferable to the other ones. It follows that at present, no definitive data in the literature support a rational choice among every-3-week docetaxel, weekly docetaxel, pemetrexed, other chemotherapeutic regimens, or EGFR inhibitors. On the basis of our metaanalysis and that of Di Maio et al,⁵⁰ for today's patients, characteristics and preferences, oncologist experience, and costs probably represent the sole criteria to select the best option for clinical decision making. It follows that further data identifying different clinical and biological profiles as predictors of different clinical response are urgently needed to overcome such a stalemate in clinical practice. In the meanwhile, docetaxel therapy every 3 weeks should still be considered the “gold standard” in the second-line treatment of metastatic NSCLC because the data about its safety in comparison with weekly docetaxel or pemetrexed are still inconclusive. Likewise, the data of efficacy and safety of all other options investigated (eg, oral topotecan, paclitaxel poliglumex, or vinflunine) did not reach a sufficient level of evidence to recommend their use in clinical practice, and those about the EGFR inhibitors probably merit further investigation to be considered a valid alternative to docetaxel in specific subgroups of patients.

Our analysis has some methodological limits. First, there is an evident methodological and clinical heterogeneity among the selected trials in the primary and secondary analysis. Such a heterogeneity is hard to be reliably assessed because of either the limited number of the selected trials (mainly in the primary analysis, where the I² or any other test to assess heterogeneity has a very low power) or the quality of the data extracted from some trials (mainly in the secondary analysis, where five trials^20–24 were available only as abstracts).

Second, the publication bias found in both the primary and secondary analyses should be kept into right account. The publication bias resulting from the low relevance given to the data from minor negative trials (that are often not published or published as minor articles) is well known. Moreover, the influence of the marketing pressure is also likely to play some role in a comprehensive analysis of the problem. Our primary analysis suggests a role of second-line treatments on the basis of a pooled analysis of three randomized trials, in which patients recruitment began in 1994 (TAX317 trial³) and ended in 2004 (ISEL trial¹⁶). The results of the pooled analysis, even though interesting, have some limits in their clinical application, and the outcome of second-line treatments (chemotherapy or EGFR inhibitors) is modest and questionable. Besides these critical considerations, the recruitment in a randomized trial comparing docetaxel therapy every 3 weeks with another chemotherapeutic regimen started in 1995,⁷ that is, just 1 year after the TAX317 trial had begun,³ and the recruitments of all the other five trials published as full text^8–10,12,13 began in 2000–2001 when the ISEL and BR21 trials^16,17 had not yet been begun and only the results of the TAX317 trial had been published. Based on these observations, it could be inferred that a sort of run-up toward fast results in second-line treatments of NSCLC may be favored by industry, more inclined to speed up the clinical use of novel molecules in a virgin field of oncology than to go deep into the real role of second-line treatments in metastatic NSCLC.

Moreover, a further “marketing bias” should be highlighted. In the last years there was an increasing habit of planning and designing “noninferiority trials” when minimal improvements in the outcomes were expected. Indeed, industry often prefers “noninferiority” trials to “superiority” trials to get easier approval from the health regulatory agencies because they require smaller series and less stringent criteria to deem results as positive or effective. Such a habit has yielded a lot of evidence about “noninferior” novel strategies, but it has also yielded considerable inconclusive data, too equivocal to be translated into clinical practice because of the fine differences in the meaning “noninferior” and “equivalent” in both the statistical and clinical lexicon. As concerns second-line treatments of NSCLC, the equivocal “fall and rise” of gefitinib is paradigmatic. The ISEL trial (a superiority trial) suggested that gefitinib was no better than placebo¹⁶; the Iressa Versus Intravenous Docetaxel (Taxotere) [INTEREST] trial (a noninferiority trial) suggested a “noninferiority” of gefitinib vs docetaxel,²³ and the TAX317 trial³ (a superiority trial) suggested that docetaxel was better than placebo.

In conclusion, second-line treatments seem to improve the outcome of NSCLC, but their actual clinical efficacy remains modest. At present, no chemotherapeutic regimen seems superior to docetaxel therapy every 3 weeks, and any alternative regimen should be carefully evaluated before treating the patients with progression after first-line chemotherapy. Moreover, impressive evidence suggests the existence of predictive factors of outcome in second-line treatments of NSCLC. A better identification and understanding of these predictive factors could represent a promising field of investigation to improve the outcome of treatments that, despite the statistical significance, have still a modest impact in clinical practice. We believe a more accurate clinical and biological selection of the patients and the identification of a “likely responder” profile are necessary to optimize our choices in daily clinical practice.^{33,34,46–49}