This randomized phase 2 study met its primary endpoint of improved PFS for the maintenance phase, based on the Cox regression model stratified by the best overall response to induction therapy (HR
0.76, one-sided p-value <0.2), indicating that continuation maintenance therapy with pemetrexed following first-line treatment with pemetrexed and cisplatin was sufficiently beneficial to warrant further investigation. The median PFS times for the maintenance phase did not differ between the two treatment arms. Although clinical researchers may be more familiar with median survival in comparing two treatment groups, this single median point comparison could be insufficient. In contrast, the HR estimate based on Cox regression model compares the whole range of survival times across the two groups; hence a more effective measure of survival difference, given the proportional hazard assumption is met. HR has been commonly used to present the primary result of survival data in oncology trials. Disease control rates, OS and one-year survival probabilities for the maintenance phase were similar for both arms. The exploratory examinations of OS for the overall study period were also similar for both arms. In addition, both the induction therapy and maintenance pemetrexed therapy were generally well-tolerated and there were relatively few grade 3/4 TEAEs observed in either study phase.
In the current study, the median PFS and OS times for the maintenance phase for the pemetrexed/BSC arm were 3.2 and 12.2
months, respectively, which are substantially shorter than the median PFS and OS times of 4.5
months and 15.5
months observed in the previous phase 3 trial in patients with advanced, non-squamous NSCLC following switch maintenance therapy with pemetrexed [5
]. It could be possible this Arab patient population may have less access to supportive care than patients from more affluent, industrialized countries, and that would favor a relatively poorer prognosis, although these statistical variations could also be explained by the smaller size of the population of the current study.
A primary consideration in designing a phase 2 clinical study is to minimize the chance that a truly active agent or regimen is erroneously rejected by keeping the probability of type II error (false-negative) low [11
]. We attempted to do this by using a higher type I error rate (alpha
0.2) than would typically be used in a larger phase 3 trial (alpha
0.05). In addition, stratified PFS was prospectively determined as the primary endpoint of the trial to minimize potential confounding from differences in response to induction therapy. This choice appears to be justified given that a higher proportion of patients in the BSC arm responded to the induction therapy compared to the pemetrexed/BSC arm (40.7% vs 25.0%). Therefore, it is possible that more patients in the BSC arm may have had a better general prognosis.
Study design limitations include the small sample size and the randomization error, which resulted in patients being randomized to the pemetrexed/BSC and BSC arms in a 1:1 ratio rather than a 2:1 ratio. Consequently, fewer patients than originally planned were randomized to the pemetrexed/BSC arm, although this did not preclude evaluation of the primary study hypothesis. These limitations may partly account for the lack of significant between-arm differences observed for the efficacy outcomes. Another potential limitation that could be worth mentioning is the absence of a placebo arm. The investigator-assessed response and disease progression are normally not precisely measured as OS is (with an exact date of death), and therefore may be subject to assessment bias, particularly in open-label studies such as this one. Obviously, we cannot rule out this possibility. However, as this is a phase 2 proof of concept study rather than a confirmatory phase 3 trial, an independent review panel was not used to validate the investigator assessments. The administration of placebos in oncology studies was reviewed by Chvetzoff and Tannock [12
], it was shown to improve symptom control (such as pain and appetite) but did not lead to tumor response. Hence, given that our primary and secondary endpoints were PFS, OS and tumor response rate, we argue that the absence of a placebo would not introduce significant bias to our efficacy outcome measures in this trial.
Recent results of a phase 3 trial provide stronger evidence of the potential benefit of continuation pemetrexed maintenance therapy [13
]. The PARAMOUNT study investigated whether pemetrexed continuation maintenance therapy would improve PFS after pemetrexed-cisplatin induction therapy in patients with advanced, non-squamous NSCLC [13
]. Randomized patients must not have progressed during the induction therapy and must have had an ECOG PS of 0 or 1 at the end of induction therapy. A total of 539 (57.4%) patients were randomized in a 2:1 ratio to continuation maintenance with pemetrexed/BSC (n
359) or placebo (normal saline)/BSC (n
180). Randomization was stratified for baseline disease stage, the best overall response to the induction therapy, and the ECOG PS just prior to randomization [13
]. The study had 90% power to show a statistically significant between-arm difference in PFS for the maintenance phase at an alpha of 0.05, assuming that the true unadjusted HR was 0.65. A median of 4
cycles of maintenance therapy was delivered in each arm. PFS from randomization was significantly longer for the pemetrexed/BSC arm than the placebo/BSC arm (median PFS: 4.1 vs 2.8
months, respectively; unadjusted HR
0.62, 95% CI: 0.49 to 0.79; p
]. Similarly, the exploratory analysis of the overall PFS (from induction) showed it was significantly longer for the pemetrexed/BSC arm than the placebo/BSC arm (median PFS: 6.9 vs 5.6
0.59, 95% CI: 0.47 to 0.74, p
0.00001). The independently reviewed disease control rate for the maintenance phase was 71.8% for the pemetrexed/BCS arm and 59.6% for the placebo/BSC arm (p
0.009). Patients achieved a median OS of 13.9
months from randomization (16.9
months from start of induction) on the pemetrexed continuation maintenance arm compared to 11.0
months from randomization (14.0
months from start of induction) on the placebo arm [14
]. The main differences in adverse events reported in the PARAMOUNT trial between the two arms were higher grade 3/4 toxicity rates for pemetrexed as follows: fatigue (4.2% vs 0.6%, respectively), anemia (4.5% vs 0.6%), and neutropenia (3.6% vs 0%) [14
]. Consistent with the safety results for our study, the safety data from the PARAMOUNT study showed that pemetrexed maintenance therapy was generally well-tolerated [5
]. The finding of an optimal maintenance therapy for patients with locally advanced or metastatic non-squamous NSCLC, who achieved disease control after first-line chemotherapy, is still the subject of study.