The role of maintenance/consolidation chemotherapy was assessed in the management of small cell lung cancer (SCLC) via a metaanalytic approach.
The role of maintenance/consolidation chemotherapy was assessed in the management of small cell lung cancer (SCLC) via a metaanalytic approach.
The Medline and Cochrane databases were searched for relevant randomized clinical trials that compared maintenance chemotherapy with follow-up. Quality of trials was assessed by European Lung Cancer Working Party (ELCWP) score. Odds ratios and rate differences were used as the effect size. Mantel–Haenszel tests with fixed and random effect models were conducted for 1- and 2-year overall survival (OAS) and progression-free survival (PFS).
Fourteen relevant randomized clinical trials to date, encompassing 2550 patients, with trial sizes ranging from 36 to 610, were identified. Both 1- and 2-year mortality were reduced with maintenance/consolidation chemotherapy. With the fixed model, odds ratios for 1- and 2-year OAS were 0.67 (95% confidence interval [CI] = 0.56–0.79), P < 0.001, and also 0.67 (95% CI = 0.53–0.86), P < 0.001. Likewise, 1- and 2-year PFS were better with maintenance/consolidation chemotherapy, with odds ratios of 0.49 (95% CI = 0.37–0.63), P < 0.001, and 0.64 (95% CI = 0.45–0.92), P < 0.015. The random model gave similar results. In accordance, maintenance chemotherapy improved 1- and 2-year OAS by 9% (from 30–39%) and 4% (from 10–14%), respectively. Similarly, 1- and 2-year PFS were also improved.
Maintenance/consolidation chemotherapy improves survival in SCLC. New randomized clinical trials are needed to further refine the place of this approach in the management of SCLC. Cancer 2005. © 2005 American Cancer Society.
Small cell lung cancer (SCLC) has a dismal outcome and only a minority of patients can be cured.1, 2 To improve this condition, a number of therapeutic approaches have been tried. Dose dense chemotherapy,3, 4 high-dose chemotherapy,5 and new molecular agents such as oblimersen (Genasense, Genta, Berkeley Heights, NJ) or imatinib mesylate6, 7 are among these approaches, with varying results. Usage of maintenance or consolidation chemotherapy has also been tried as a means of improving survival in SCLC, but reviews on this topic generally concluded that maintenance chemotherapy was of no proven benefit.8, 9 Indeed, only a few randomized clinical trials (RCTs) reported so far (3 of 14) have shown an overall survival difference in favor of maintenance/consolidation chemotherapy.10–12
Moreover, these trials have been regarded as heterogeneous in terms of their design and to have employed different chemotherapy regimens, making it difficult to draw clearer conclusions.13 However, given the doubt over the efficacy of maintenance/consolidation chemotherapy in SCLC, and the probability that a large number of patients could potentially be better served with this approach, we decided to conduct a metaanalysis of the published literature to understand whether this approach could have any promise in the management of SCLC.
The PubMed and Cochrane databases were searched for RCTs that performed a direct comparison between a maintenance/consolidation regimen and a no-treatment (follow-up) arm after the completion of induction therapy. For practical purposes, we did not make a distinction between maintenance and consolidation approaches, as these were not clearly described in previous studies, and as we primarily wanted to explore the effect of continuation of chemotherapy beyond the standard 4 to 6 cycles. Thus, among the inclusion criteria were 4 to 6 cycles of induction chemotherapy, and also more than 6 cycles of chemotherapy in total for the induction plus maintenance/consolidation phases. As explained above, this was because 4 to 6 cycles of induction treatment is the standard today,14, 15 and with any further maintenance/consolidation treatment the total number of cycles administered should therefore add up to more than 6, by definition.
The search was limited to RCTs published in English after 1966. The keywords maintenance chemotherapy, consolidation chemotherapy, lung cancer, small cell carcinoma, and oat cell carcinoma, combined with appropriate Boolean operators, were used for the search strategy. Additionally, references from relevant reviews and textbooks in this field were evaluated. Two medical oncologists (H.B. and M.A.) first independently and then collaboratively reviewed potential abstracts and selected full text articles of relevant trials for further review. For classification purposes, maintenance/consolidation chemotherapy was accepted as “same as induction,” if all the chemotherapeutics included in maintenance had also been used in the induction regimen.
Quality of trials was assessed by European Lung Cancer Working Party (ELCWP) score, referring directly to previously published ELCWP scores. If not published previously, scores were calculated according to the published guidelines.13
The odds ratio was used to estimate the risk of mortality or disease progression at certain time points: 1- and 2-year overall survival (OAS), 1- and 2-year progression-free survival (PFS) figures from the relevant articles were abstracted. If these measures were already stated in the article, surviving proportions and patient numbers were directly used. Conversely, in most cases where these measures were not directly given in the article, they were abstracted from the survival curves. The value for the proportions surviving or surviving progression-free at certain time points were multiplied by the total number of cases in that arm to find the total number of overall and progression-free survivors. Despite the fact that this approach is an approximation, it has been accepted as reasonably accurate, and thus has been used in the literature.16, 17 For analyzing the effect of maintenance/consolidation chemotherapy on PFS, only those articles with the required information on PFS were used. The Mantel–Haenszel test18 with fixed and random effect models was conducted to analyze the effect of maintenance/consolidation chemotherapy on OAS and PFS. In addition, rate differences for PFS and OAS at 1 and 2 years were also calculated by Mantel–Haenszel test with fixed and random effect models. A P value < 0.05 was considered significant.
Heterogeneity of trials was evaluated by χ2 based Q tests. In addition, Forest plots were constructed. Publication bias was assessed by Funnel plots.
In addition, metaregression analyses were conducted to explain the variations in effect size. To do this, logarithmic transformation of the odds ratio was used as the dependent variable.19 Prognostic factors such as proportion of patients with complete response to induction regimen and proportion of those with limited disease at the onset of trial, as well as other features like type of induction therapy (cisplatin-containing vs. others), type of maintenance/consolidation regimen (same as induction vs. others), year of trial onset, ELCWP scores, and size of trial (logarithmic transformation) were entered into linear regression analysis.
Comprehensive Metaanalysis v. 1.0.25 (Biostat, Inc; Englewood, NJ) was used for this metaanalysis. In addition, metaregression analysis was conducted by SPSS (Chicago, IL; release 11.0). For the purpose of better reporting the results of this metaanalysis, we adhered to “The Quality of Reporting of Metaanalyses” (QUOROM) statement.20
Eighty-four randomized clinical trials (RCTs) were evaluated at the initial stage. A total of 14 RCTs met the inclusion criteria.10–12, 21–31 (See Fig. 1 [flow diagram] for the exclusion criteria of various trials.) A study that was evaluated for the role of maintenance chemotherapy in a previous review was not included in this metaanalysis, primarily because it did not involve maintenance and no-maintenance arms.32 The total number of all patients in this metaanalysis was 2550, and all were suitable for OAS analysis. However, only 8 trials had data for PFS, encompassing 1727 patients. Three trials were identified as superior10–12 and one trial inferior OAS27 with the maintenance approach, with statistical significance. Six studies also found better PFS with maintenance/consolidation chemotherapy.10, 21, 22–24, 26 (See Table 1 for characteristics of RCTs included in this metaanalysis.) Funnel plots did not indicate publication bias.
|Citation||Onset of trial||N Totala||ELCWP scoreb||Induction||Maintenance||Stagec||Statusd||Outcome||Toxicity|
|Maurer et al. (1980) 31||1972||258/ 36||39/ 102||6 × (C+Mtx, or C+Mtx+V or C+Mtx+V+Lv) +/- RT vs. 6 × C +/- RT||same with induction on every-other month basis until relapse||LD+ED||CR||Median OAS = 16.8 months with maintenance and 6.8 months without maintenance (P = 0.01)e||Not seperately reported for the maintenance arm|
|Cullen et al. (1986) 25||1980||309/ 93||37/ 102||6 × CAV +/- RT||8 × CAV at lower dosage (every 4 weeks)||LD+ED||CR (+GR)||Median OAS increased in extensive disease, no change in limited disease (P = 0.006, and > 0.05)e||Only 2 patients withdrawn by clinicians due to toxicity in the maintenance arm|
|Einhorn et al. (1988) 10||1982||160/ 148||67/ 102||6 × CAV +/- RT||4 × EP||LD||CR+PR||Median OAS = 78.3 weeks with maintenance and 49 weeks without maintenance (P = 0.01)fPFS also better with maintenance (P< 0.001)||No severe toxicity after maintenance|
|Spiro et al. (1989) 11||1982||610/ 610||69/ 102||4 × CEV +/- RT||4 × CEV||LD+ED||CR+PR+SD||Median OAS with 4 courses alone inferior to 8 courses or to 2nd line chemotherapy at progression (Pf. No direct comparison for PFS.||11 toxic deaths in the maintenance arm vs. 18 in the 4 courses arm.|
|Byrne et al. (1989) 27||1981||68/ 66||60/ 102||3 × (EP 3 weekly+ CVM 4 weekly) + RT (total of 6 courses)||6 × CVM||LD||CR+PR+SD||Median OAS with maintenance is inferior (14.1 vs. 19.2 months, P = 0.05)f||With maintenance, 5 patients developed leucopenia < 2000/mm3.|
|Bleehen et al. (1989) 30||1981||497/ 265||43/ 102||6 × ECVM +/- RT||6 × ECVM||LD+ED||CR+PR||No median OAS difference with maintenance (P = 0.27), if CR to induction, median OAS different (P < 0.05)e||All drugs stopped in 12% at induction, and in 24% at maintenance phase|
|Ettinger et al. (1990) 28||1982||577/ 86||42/ 102||3-4 × CAV-HEM vs. 6-8 × CAV||(CAV-HEM or CAV) for a maximum of 84 weeks||ED||CR||CAV-HEM maintenance detrimental (P = 0.13), CAV maintenance useful (P = 0.09)a. PFS better with maintenance for CAV treated cases (P = 0.015).||42% recieving maintenance had life threatening toxicity|
|Lebeau et al. (1991) 29||1983||320/ 79||64/ 102||CAEL 4 weekly until CR or progression +/- RT||6 × CAEL||LD+ED||CR||Median OAS 332 days with maintenance and 246 days without maintenance (P = 0.41)f||No data|
|Johnson et al. (1993) 12||1982||386/ 151||67/ 102||6 × CAV + RT||2 × EP||LD||CR+PR||Median OAS 21.1 months with maintenance and 13.2 months without maintenance (P = 0.028)f||At least 1 grade 4 toxicity:73% vs. 56% with and without maintenance (P = 0.002)|
|Giaccone et al. (1993) 26||1982||687/ 434||67/ 102||5 × CAE 3 weekly +/- RT||7 × CAE||LD+ED||CR+PR+SD||No OAS differencef. Median PFS (177 vs. 114 days) improved with maintenance (P = 0.0004)||13 toxic deaths during induction and 3 at maintenance|
|Sculier et al. (1996) 24||1990||235/ 91||85/ 102||6 × IEA or IEEp 3 weekly||12 × EVi||LD+ED||CR+PR||Median OAS 48 and 38 weeks with and without maintenance (P = 0.10)f, Median PFS penia, 2 grade 2 neuropathy 25 vs. 12 weeks (P = 0.003)||Maintenance well tolerated; 1 grade 4, 9 grade 3 lueuco|
|Beith et al. (1996) 23||1981||202/ 124||61/ 102||4 × EP +/- RT||10 × CAV||LD+ED||CR+PR||Median OAS 54 and 52 weeks with and without maintenance (P = 0.72)b, Median PFS 37 vs. 23 weeks (P = 0.099)||Three deaths due to neutropenic sepsis in the maintenance arm.|
|Schiller et al. (2001) 22||1995||402/ 223||73/ 102||4 × EP||4 × Topotecan||ED||CR+PR+SD||Median OAS 9.3 vs. 8.9 months between maintenance and obsernvation arms (P = 0.43)f Median PFS 3.6 vs. 2.3 months (P||Grade 4 neutropenia and 60% vs.50% in maintenance and observation arms.|
|Hanna et al. (2002) 21||1993||233/ 144||63/ 102||4 × IEP||3 × oral E (1 cycle = 3 weeks therapy, 1 week off)||ED||CR+PR+SD||Median OAS 12.2 and 11.2 weeks with and without maintenance (P = 0.0704)f, Median PFS 8.2 vs. 6.5 weeks (P = 0.0018)||Grade 4 granulocytopenia in 12 of 72 maintenance patients|
The fixed model showed significantly less 1- and 2-year mortality with maintenance chemotherapy. The effect size (odds ratio) for 1-year OAS was 0.67 (95% confidence interval [CI] = 0.56–0.79), P < 0.001, and for 2-year OAS, also 0.67 (95% CI = 0.53–0.86), P < 0.001. Likewise, with the random model the effect size for 1- and 2-year OAS were 0.65 (95% CI = 0.53–0.80), P < 0.001, and 0.68 (95% CI = 0.49–0.93), P = 0.016, respectively. This corresponded to an increase of 9% in 1-year OAS (from 30–39%), and 4% in 2-year OAS (from 10–14%), P < 0.05, in both fixed and random models. Tests for heterogeneity were not significant for 1- and 2-year OAS (see Figs. 2, 3 for forest plots).
The fixed and random models both showed improved outcome with maintenance chemotherapy for 1- and 2-year PFS. The odds ratios for fixed and random models were, respectively, 0.49 (95% CI = 0.37–0.63), P < 0.001, and 0.47 (95% CI = 0.34–0.66), P < 0.001, for 1-year PFS, and 0.64 (95% CI = 0.45–0.92), P = 0.015, and 0.65 (95% CI = 0.39–1.09), P = 0.099 for 2-year PFS. Tests for heterogeneity were not significant for 1- and 2-year PFS. In terms of improvement in PFS by maintenance chemotherapy, 1-year PFS increased by 10% (from 13–23%), and 2-year PFS by 3% (from 10–13%); P < 0.05 for 1-year PFS by both models. However, for rate differences of PFS at 2 years, the fixed model yielded a significant result (P < 0.05), whereas the random model did not (P > 0.05) (see Figs. 4, 5 for forest plots of PFS).
Of the independent variables entered into the regression analysis, i.e., proportion of patients with complete response to induction regimen, proportion of patients with limited disease at the onset of trial, type of induction therapy (cisplatin-containing vs. others), type of maintenance/consolidation regimen (same as induction vs. others), year of trial onset, ELCWP scores, and size of trial (logarithmic transformation), none were found to be significantly associated with the log odds ratio for the four different types of outcome (1- and 2-year OAS, 1- and 2-year PFS).
The main finding from this metaanalysis is that maintenance chemotherapy decreases the 1- and 2-year risk of death or disease progression in SCLC. Indeed, maintenance chemotherapy decreases the odds of 1- and 2-year death by a third, or, in other words, improves 1- and 2-year OAS by 10% and 4%, respectively, in patients with SCLC. Likewise, with maintenance chemotherapy the odds of progression and/or death is reduced by 51% and 46%, respectively, 1 or 2 years after the onset of maintenance chemotherapy. Previously, as far as direct comparisons between maintenance and follow-up arms are concerned, three studies had shown improved OAS and five studies had found better PFS with maintenance/consolidation chemotherapy. However, the majority of randomized controlled trials had not indicated significantly superior OAS or PFS with this approach. Accordingly, the National Cancer Institute's PDQ14 summaries and ESMO15 minimum clinical recommendations still advise 3 to 6 months, or 4 to 6 cycles, of chemotherapy, and this approach is regarded as the “gold standard” today. In this regard, this metaanalysis challenges this view. In parallel to what we have found, some authors in the field of oncology feel that maintenance chemotherapy may be beneficial at least for a subset of SCLC patients.33 This metaanalysis supports this idea.
However, some of the trials showed increased toxicity with the use of maintenance regimens. Notably, three of the trials in this metaanalysis12, 23, 28 indicated that the benefit and toxicity from this approach should be balanced.
Our metaregression analysis shows that the effect of maintenance chemotherapy on OAS and PFS was independent of various patient or trial characteristics. In other words, the features of patient cohorts or trials included in this metaanalysis were not associated with the effect size of maintenance chemotherapy. This lack of association shows at the same time that the benefit of maintenance chemotherapy is consistent across various patient groups as evaluated in this analysis. However, it may also suggest that this analysis did not evaluate other relevant variables, known or unknown, in order to describe which patients do or do not benefit.
It is interesting that in two trials showing the superiority of maintenance chemotherapy,10, 12 the induction therapy was cyclophosphamide, adriamycin, and vincristine (CAV) and the maintenance regimen was etoposide and cisplatin (EP). Although our metaregression analysis shows that the types of induction and maintenance regimens are not related to the effect size, these two trials perhaps show that the inclusion of a less cross-resistant second-line chemotherapeutic regimen might improve the therapeutic efficacy of the whole strategy.
Unfortunately, only four of the RCTs in this metaanalysis provide information on treatment at relapse.11, 26, 29, 30 Notably, Spiro et al.11 showed that if additional treatment was given at relapse, the use of maintenance chemotherapy was not associated with superior survival. However, these four trials also demonstrated that, for various reasons, a large portion of SCLC patients could not be offered treatment at relapse. Thus, taken together, these findings possibly suggest that some benefit of maintenance chemotherapy is due to earlier administration of additional chemotherapy after induction regimens, when chemotherapy can still be administered in fitter patients prior to progression of the disease.
Previous reviews on the role of maintenance chemotherapy concluded that a metaanalysis in this field would not be feasible, primarily because the trials were heterogeneous and summary data were difficult to abstract. Indeed, trials included in this metaanalysis showed a wide variation in size or the treatment regimens used. However, inspection of the forest plots and the results of Q statistics indicated that this heterogeneity did not reach the level of significance. Nevertheless, because of previous suspicion concerning the heterogeneity of trials, we also chose to conduct a random effect analysis, which again showed that patients receiving maintenance chemotherapy did significantly better. In addition, abstracting the event rates from the survival curves was possible for calculation of the odds ratio. However, as this is not a metaanalysis of individual patient data, our results should be considered with caution.
As stated above, this metaanalysis shows that maintenance/consolidation chemotherapy is active in SCLC, although it is of marginal benefit in some of the individual studies. This discrepancy is caused to a great extent by the fact that in the majority of cases, a single study simply does not have the sample size required to precisely estimate the impact of a treatment. Indeed, in the literature there are metaanalyses with sizes similar to ours, where the vast majority of individual studies do not yield any difference for a treatment, whereas the result of the metaanalysis is clinically and statistically significant. For example, in a metaanalysis by D'addario et al.,34 the authors showed that in patients with advanced nonsmall cell lung cancer in 14 trials using third-generation agent-containing combination regimens (with or without platinum), the objective response rate was significantly improved by platinum arms as compared to nonplatinum arms, although only 1 of these 14 individual trials yielded significantly better objective response rates in favor of the platinum arm.
As this metaanalysis does not and cannot answer the questions of whom to treat and with what to treat, properly designed and large clinical trials stratifying patients into prognostic groups and evaluating different maintenance regimens will be required to answer these important issues. Nevertheless, we provide the first metaanalysis that shows the superiority of maintenance or consolidation chemotherapy over follow-up after standard therapy. If confirmed, this finding will transform the way we treat SCLC. In short, this metaanalysis suggests that maintenance/consolidation chemotherapy should play a role in the management of SCLC. We call for future work in the form of new, large, randomized, controlled clinical trials to support these findings.
The authors thank Aventis for providing some of the key articles.