Secondary analyses of the randomized phase III Stop&Go study: efficacy of second-line intermittent versus continuous chemotherapy in HER2-negative advanced breast cancer

Abstract Background: Previously, we showed that reintroduction of the same (first-line) chemotherapy at progression could only partially make up for the loss in efficacy as compared to continuously delivered first-line chemotherapy. Here, we report the probability of starting second-line study chemotherapy in the Stop&Go trial, and the progression-free survival (PFS) and overall survival (OS) of patients who received both the first- and second-line treatment in an intermittent versus continuous schedule. Methods: First-line chemotherapy comprised paclitaxel plus bevacizumab, second-line capecitabine or non-pegylated liposomal doxorubicin, given per treatment line as two times four cycles (intermittent) or as eight consecutive cycles (continuous). Results: Of the 420 patients who started first-line treatment within the Stop&Go trial (210:210), a total of 270 patients continued on second-line study treatment (64% of all), which consisted of capecitabine in 201 patients and of non-pegylated liposomal doxorubicin in 69 patients, evenly distributed between the treatment arms. Median PFS was 3.7 versus 5.0 months (HR 1.07; 95% CI: 0.82–1.38) and median OS 10.9 versus 12.4 months (HR 1.27; 95% CI: 0.98–1.66) for intermittent versus continuous second-line chemotherapy. Second-line PFS was positively influenced by prior hormonal therapy for metastatic disease and longer first-line PFS duration, while triple-negative tumor status had a negative influence. Patients with a shorter time to progression (TTP) in first-line (≤10 months) had a higher probability of starting second-line treatment if they received intermittent compared to continuous chemotherapy (OR 1.97; 95% CI: 1.02–3.80). Conclusion: We recommend continuous scheduling of both the first- and second-line chemotherapy for advanced breast cancer.


Background
Although several improvements have been made in the treatment of early breast cancer, advanced breast cancer remains largely incurable with a median survival of $2-3 years [1,2]. Treatment of patients with advanced breast cancer focuses both on improving length and quality of life, but the optimal duration of a particular line of chemotherapy is still not clarified [3]. This results in a high variety of treatment strategies in daily clinical practice, with some using a predefined number of cycles and others continuing treatment till progression of disease (PD) or unacceptable toxicity. Considering the fact that the likelihood of response to each subsequent line of treatment decreases, optimization of especially the initial treatment lines may be important.
A previous meta-analysis of 11 randomized trials on shorter (4-6 cycles) versus longer (8-12 cycles) chemotherapy durations reported inferior PFS results with shorter durations of treatment [4]. However, most of the trials included in this pooled analysis comprised outdated treatment agents, with only three trials investigating taxane-containing treatments, showing inconclusive results [5][6][7]. As taxanes are suggested for first-line chemotherapy by current guidelines [3], the question remained if these recommendations were also applicable for these modern agents, with expected higher efficacy in the initial treatment cycles. Therefore, the Stop&Go study intended to make a comparison of an interrupted versus a continued chemotherapy schedule with agents that were frequently used in current clinical practice.
The Stop&Go trial is a phase III trial, investigating the efficacy and tolerability of intermittent versus continuous scheduling of the first two chemotherapy lines [8]. For first-line chemotherapy [8], we found that reintroduction of the same therapy at progression (intermittent schedule) could only partially make up for the loss in efficacy as compared to continuously delivered chemotherapy. In the current exploratory analysis, we focus on the results of second-line treatment scheduling. We hypothesized that the most efficient treatment approach in first-line (intermittent or continuous) would lead to a differential patient selection in second-line and to an increased effectiveness in second-line and overall, when using the same approach in second-line.

Study design and participants
The randomized, open-label phase III Stop&Go trial was conducted at hospitals affiliated to the Dutch Breast Cancer Research Group (BOOG). Patients with HER2-negative incurable locally advanced or metastatic breast cancer who were candidates for first-line chemotherapy and who fulfilled the eligibility criteria were randomized to intermittent or continuous chemotherapy [8]. Participants were allocated in a 1:1 ratio by minimization, using the computer program ALEA (https://www.tenalea.com/nkiavl/alea/Default.aspx). Stratification factors included institute, hormone receptor status and site of disease (visceral versus non-visceral). No placebo treatments were used. After allocation, participants were to receive the same treatment strategy (intermittent or continuous) both in first-and in second-line (no cross-over and no repeated randomization).
The study was conducted in agreement with the Declaration of Helsinki (version 1 May 1996) and local regulations (ethics committee Eindhoven, the Netherlands). All participants provided written informed consent. The study protocol is registered at the EU Clinical Trials Register, number 2010-021519-18 (https://www.clinicaltrialsregister.eu/ctrsearch/search?query=2010-021519-18).
In the intermittent treatment arm, four cycles of chemotherapy were given followed by another four cycles of the same treatment, if a PD-event occurred !3 months after the initial cycles. If PD occurred <3 months or after a second PD after reintroduction, second-line treatment had to be started if the patient was still fit enough. In the continuous treatment arm, a maximum of eight cycles of chemotherapy were given successively in both first-and second-line (Figure 1), without re-introduction at PD. In the current analysis, we determined the effect of chemotherapy scheduling in the patients who were able to receive both first-and second-line study treatment.

Treatment and assessments
First-line treatment consisted of paclitaxel 90 mg/m 2 intravenously on days 1, 8 and 15, combined with bevacizumab 10 mg/kg intravenously on days 1 and 15, repeated every 28 days. Bevacizumab maintenance was continued at 15 mg/ kg once every 21 days until the occurrence of PD or unacceptable toxicity. For extensive information, see previous publication [8]. Second-line treatment consisted of either capecitabine at 1000 mg/m 2 orally BID for 2 weeks, followed by 1 week off, or of non-pegylated liposomal doxorubicin intravenously at 60 mg/m 2 once every 3 weeks, chosen at the discretion of the physician and depending on prior use of anthracyclines in the adjuvant setting. Receiving endocrineor other systemic anti-tumor treatmentsduring the study period was considered a major protocol violation, leading to censoring within the PFS analyses.
Toxicity-related dose adjustment or treatment delays were to be performed according to dose-modification guidelines described in the study protocol (see Online Supplement).
Disease response assessment was based on local clinical evaluation according to the RECIST criteria version 1.1 [9]. During the follow-up period, date of progression after the end of study treatment, and start date of new anti-tumor treatment was registered every three months until death or study withdrawal.

Outcomes
Here, we report on secondary endpoints from the Stop&Go study, with the main endpoint being PFS of second-line treatment, calculated from the start of second-line study treatment until the date of final progression as defined by the investigator, or death, whichever occurred first. Additionally, the combined PFS of both first-and second-line treatment was assessed, from randomization date to date of progression after two study-line treatments, or date of progression on first-line treatment if not started with second-line, or death, whichever occurred first. Within the PFS analyses, patients who started non-protocol anti-tumor treatment before PD were censored at the date the non-protocol anti-tumor therapy was started or at the date of last follow-up if non-progressive. OS was calculated from the start of second-line study treatment as well as from randomization. Other endpoints included predictive factors for PFS in second-line and the probability of starting second-line treatment. For the probability of starting secondline treatment, patients who died during first-line treatment were excluded. Predictive parameters included the time to progression (TTP) in first-line treatment, defined as the time from randomization until date of final progression defined by the investigator. Toxicity of second-line treatment was graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) version 4.0. All AE's grade !3 were recorded.

Statistical analyses
Sample size calculations were done based on the primary endpoint of the Stop&Go study, first-line PFS. In order to demonstrate non-inferiority of the intermittent arm compared to the continuous arm for this endpoint with 80% power at a 0.025 significance level, a total number of 420 patients were required for randomization to first-line treatment [8]. All efficacy analyses, comparing the intermittent with the continuous treatment arm, were performed according to the intention-to-treat principle. The safety analysis included all patients who received at least one dosage of second-line study medication. PFS and OS endpoints measured from randomization were performed on all patients, and separately for those who started second-line treatment. In this paper, we only report results for the patients who received both the first-and second-line study treatment. Results for all randomized patients are displayed in the online supplement, and include PFS and OS endpoints of the combined first-and second-line treatment (Online Figure 2).
The Kaplan-Meier (KM) method was used to assess PFS and OS, calculated from start of second-line study treatment and from the date of randomization. Survival curves were compared between study arms by log-rank tests. To account for possible unbalanced compliance in the study arms with respect to second-line treatment, a propensity score (PS) model was estimated, and stabilized inverse probability weights (SIPW) were calculated. Curves weighted by SIPWs were also displayed and differences between treatment groups were assessed. The (weighted) Cox proportional-hazards model was used to estimate hazard ratios (HRs) of the intermittent arm versus the continuous arm and their corresponding 95% confidence intervals (CIs). The Cox model was stratified by hormonal receptor status and site of disease, and adjusted for other patient and clinical characteristics in the case of multivariable analyses. The proportional hazards assumption was assessed by including time-by-covariate interaction terms in the model and scaled Schoenfeld residuals plots over time. In case the hazard ratio changed over time, Cox models were fitted for different time periods where the assumption of proportionality might hold. No multiple imputation of missing values in clinical characteristics was performed for multivariable analyses, as the percentage missing was 5% or less. Relevant multivariable analyses are displayed in the full paper. The remaining results, including weighted analyses, are displayed in the Online Supplement (Online Figures 2 and 3).
To model the probability of starting second-line study treatment, a mixed-effects logistic regression was performed among eligible patients (i.e., those alive after the end of firstline treatment), adjusting for several patient and clinical characteristics.

Patient characteristics and received treatments
Of the 420 patients who started first-line treatment within the Stop&Go trial (210:210), a total of 270 patients continued on second-line study treatment (64% of all), which consisted of capecitabine in 201 patients and of non-pegylated liposomal doxorubicin in 69 patients, evenly distributed between the treatment arms (Online Figure 1). The key characteristics at randomization were well-balanced between treatment arms for the patients who started second-line study treatment, and these were in line with the baseline characteristics of the total study population ( Table 1).
Main reasons for final second-line treatment discontinuation in the safety population were disease progression (55% in both arms), major protocol violations (e.g., receiving more than the prescribed amount of cycles; 24% versus 0%) and toxicity (11% versus 10%), for the intermittent and the continuous arms, respectively (Online Table 2).

PFS in second-line and combined PFS of first-and second-line
Median PFS on second-line study treatment (n ¼ 270) was 3.7 months (95% CI: 2.8-4.7) for the intermittent treatment arm versus 5.0 months (95% CI: 4.4-5.9) for the continuous treatment arm, with a HR of 1.07 (95% CI: 0.82-1.38) in a Cox model stratified by hormonal receptor status and site of disease ( Figure 2(A)). However, the assumption of proportionality of hazards was not met, so a time-by-arm interaction term was introduced. The HR was found to decrease with time. Period-specific HRs indicated that until 7 months after the start of second-line treatment, the risk of progression was higher in the intermittent arm (HR 1.39; 95% CI: 1.01-1.92, adjusted for hormonal receptor status, site of disease and other patient and clinical characteristics available in 255 patients). Considering the small population of patients left at risk at 7 months, caution is required in comparing the treatment arms beyond this point.
For the 270 patients starting second line of treatment, the combined median PFS of first-and second-line treatment was 14.6 months versus 16.4 months for intermittent versus continuous chemotherapy scheduling, with a HR of 1.12 (95% CI: 0.86-1.45) (Figure 2(C)). However, the same phenomenon was seen as for the PFS of second-line alone: non-proportional hazards were found for study treatment. At 12 months after randomization, the PFS curve reached a small plateau for the intermittent arm. Up until this time, hazards for combined PFS were significantly higher with intermittent treatment (HR 1.85; 95% CI: 1.14-3.00 adjusted for hormonal receptor status, site of disease and other patient and clinical characteristics available in 255 patients).   study treatment was 0.90 (95% CI: 0.59-1.37). The odds of starting second-line treatment were lower for patients with baseline ECOG performance status 2 compared to 0-1 (OR 0.24; 95% CI: 0.09-0.63) ( Table 3). The probability of starting second-line treatment in the different treatment arms was linked to the TTP in first-line. Within the continuous arm, the probabilities of starting second-line study treatment were 0.80 and 0.61 for patients with a first-line TTP of >10 months and 10 months, respectively. For the intermittent treatment, these probabilities were 0.68 and 0.75. To further evaluate this association, we first looked at the treatment arms separately. For the continuous arm, the odds of starting second-line chemotherapy in patients with TTP >10 months were significantly higher compared to the odds for patients with a TTP of 10 months (adjusted OR 2.73; 95% CI: 1. 36-5.46). In contrast, no evidence for such interaction was found for the intermittent treatment arm (OR 0.70; 95% CI: 0.34-1.43). Secondly, we compared the odds of starting second-line study treatment between treatment arms for patients with a specific TTP duration in first-line. Patients with a TTP of 10 months had a higher probability of starting second-line treatment if they received intermittent treatment compared to continuous treatment (OR 1.97; 95% CI: 1.02-3.80). For a TTP of >10 months, there was an indication of the opposite (OR 0.50; 95% CI: 0.24-1.06). See Table 3 for all parameters.

Discussion
The Stop&Go study was designed to assess the impact of chemotherapy scheduling in first-and second-line chemotherapy in patients with HER2-negative advanced breast cancer. Previously, we reported that continuous chemotherapy during first-line treatment might be the preferred strategy, because of the observed trend in improved PFS and OS as compared to intermittently delivered first-line chemotherapy [8]. Now, we addressed the question whether maintaining the same scheduling during second-line would further improve outcome. Indeed, we found that continuous scheduling of both first-and second-line treatment showed an improved OS by approximately two months (21.0 versus 23.2 months, at a HR of 1.27 [95% CI: 0.98-1.66]). Although there was an absence of statistical significance, results showed a clear favorable trend for continuous treatment.
Although the observed efficacy of second-line study treatment (median PFS 3.7 versus 5.0 months, median OS 10.9 versus 12.4 months) seemed relatively poor, results were comparable to other studies. A review on phase II and III studies on single-agent second-line chemotherapy for advanced breast cancer noted median OS of 8-13 months for the majority of trials [10]. In comparison, trials reporting second-line PFS within this review were scarce, and medians varied from 2.5 to 9.8 months depending on the investigated single-agent [10]. Observational studies on multiple lines of chemotherapy for advanced breast cancer reported median second-line PFS or TTP between 2.5 and 11.7 months [11][12][13][14][15][16][17][18][19][20][21]. However, comparison of these outcomes between trials is limited due to the large variety in (amount of) previous treatments, selection of patients, type of chemotherapy and whether or not concurrent treatments were given.
To our knowledge, this is the first randomized controlled trial that evaluated the efficacy of intermittently delivered chemotherapy over several treatment lines. Although other studies comparing different durations of first-line chemotherapy found significant benefits in PFS with prolonged chemotherapy, effects on OS were less consistent [5,6,[22][23][24][25][26][27][28][29][30][31][32]. From this current study, we learned that the same phenomenon occurred in first-and second-line study treatment; a relatively rapid progression of disease in the period after cessation of chemotherapy within the intermittent arm compared to a more stable course of disease within the continuous treatment arm. Additionally, the exploratory multivariable Coxregression analysis indicated that the duration of PFS in firstline was of significant influence on second-line PFS (Table 2).
To implement a continuous chemotherapy schedule in clinical practice, the challenge is to use agents that provide disease control, are well tolerated and can be continued for a prolonged period without interruptions. Several studies have investigated the use of low-dose metronomic chemotherapy schedules, indicating the ideal agents should preferable be oral, have limited cumulative toxicity and low costs [33]. Possible candidates for the treatment of advanced breast cancer from phase I and II trials include oral cyclophosphamide and methotrexate, capecitabine and oral vinorelbine. [33] Additionally, the effects of metronomic chemotherapy include immune-mediated as well as endocrine, anti-angiogenic and stroma-targeted mechanisms, with several studies in advanced breast cancer suggesting a possible synergistic effect of combining metronomic chemotherapy with endocrine, immune-and/or targeted therapies [34][35][36][37][38][39][40][41]. Future research should thus focus on continuous administration of well-tolerated chemotherapy agents, as a background for other therapies. Additionally, the

Infections and infestations
Urinary tract or bladder infection -  Pain included AEs labeled pain, back pain, abdominal pain, pelvic pain, thoracic pain, Bone pain, pain in extremity, headache, non-cardiac chest pain, pleuritic pain and hepatic pain. Cardiac chest pain was not included. c Additional AEs of grade 3 or higher of special interest occurred in the following incidences, taking into account the maximum grade per patient: heart failure/ left ventricular systolic dysfunction/decreased ejection fraction n ¼ 1 versus n ¼ 3, myocardial infarction n ¼ 0 versus n ¼ 1, small intestinal perforation n ¼ 1 versus n ¼ 0, DPD-deficiency n ¼ 1 versus n ¼ 0, for the intermittent versus the continuous arm, respectively. implementation of continuous chemotherapy requires us to rethink the optimal sequence of available agents. Possibly, the more tolerable agents should be exploited first before moving onto the more toxic agents that can only be given for a limited amount of time.
A note of caution is due, since the lack of significant differences in second-line PFS might be caused by the fact that 24% of patients in the intermittent treatment arm received more than the allocated number of subsequent chemotherapy cycles (major protocol violation). If these patients would have been excluded from current analysis (a per protocol instead of intent-to-treat analysis), differences between the treatment arms could have been greater in disadvantage of intermittent scheduling. Additionally, more patients were censored due to start of non-protocol treatment (e.g., hormonal therapy) in the intermittent arm (n ¼ 24 versus n ¼ 16). Hypothetically these protocol violations might be due to uncertainty about the outcome caused by the chemotherapy-holiday, a relevant factor to take into account as endocrine consolidation treatment is a common practice. Another limitation includes the inability to soundly compare the subgroups of patients that received capecitabine as second-line study treatment (n ¼ 201) with those who receives non-pegylated liposomal doxorubicin (n ¼ 69), and the populations of patients in second-line with hormonereceptor positive (n ¼ 220) disease with those who had triple-negative disease (n ¼ 40) due to the imbalanced numbers. In the Stop&Go study protocol, the prolonged treatment arm consisted of eight treatment cycles. As the toxicity of especially capecitabine is generally mild, we now would recommend to continue therapy until progression of disease if well-tolerated. But, based on the results of our study we cannot make that recommendation as we did not formally test the benefit of more than eight treatment cycles.
In conclusion, the Stop&Go study found no efficacy benefits from an intermittent chemotherapy schedule across two treatment lines. The superior period-specific progression-free survival and the advantageous overall survival results of the continuous schedule prompt our advice to schedule both the first-and second-line chemotherapy for patients with HER2-negative advanced breast cancer without interruptions.