Anti-EGFR-based maintenance versus stop and go in patients with left-sided, non-MSI-H, RAS/BRAF-wt metastatic colorectal cancer: individual patient data pooled analysis☆ Authors: A. Raimondi, G. Tinè, V. Boige, A. De Stefano, A. Ballhausen, S. Gourgou, F. Giuliani, A. Stahler, G. Randon, M. Prisciandaro, P. Manca, M. Ambrosini, R. Fazio, S. Mohamed, G. Gronchi, G. Santabarbara, E. Brument, A.H.S. Alig, R. Miceli, O. Bouche, D.P. Modest, A. Avallone, F. Pietrantonio Published: 2026 ABSTRACT BackgroundDoublet chemotherapy plus an anti- epidermal growth factor receptor (EGFR) agent is a standard of care in patients with left-sided, microsatellite stable, RAS/BRAF wild-type (wt) mCRC. No conclusive evidence is available to guide treatment de-escalation, but several randomized clinical trials investigated either maintenance or intermittent treatment schedules (stop and go).Patients and methodsWe performed an individual patient data pooled analysis of four multicenter, randomized phase II trials (PanaMa, Valentino, PRODIGE-28 TIME, IMPROVE). Only patients with left-sided, non-microsatellite instability-high, RAS/BRAFV600E-wt who started the protocol-planned post-induction treatment were included and stratified into three treatment groups: maintenance with 5-fluorouracil/leucovorin (5-FU/LV) plus anti-EGFR, maintenance with anti-EGFR alone, or stop and go. The primary endpoint was overall survival (OS) with stop and go versus 5-FU/LV plus anti-EGFR maintenance. Secondary endpoints were progression-free survival (PFS) and time to failure of strategy (TFS) with stop and go versus 5-FU/LV plus anti-EGFR; and OS, PFS, and TFS with stop and go versus anti-EGFR alone.ResultsA total of 378 molecularly selected left-sided patients were included. Overall, 166, 109, and 103 patients received maintenance with 5-FU/LV plus anti-EGFR, anti-EGFR alone, or stop and go, respectively. No significant differences in OS were reported between stop and go versus 5-FU/LV plus anti-EGFR maintenance [median 26.3 versus 29.7 months, hazard ratio (HR) 1.24, P = 0.530], or stop and go compared with anti-EGFR alone (median 26.3 versus 30.1 months, HR 1.39, P = 0.165). PFS was significantly shorter with stop and go, while TFS was comparable between stop and go compared with both maintenance schedules.ConclusionsThis pooled analysis of candidates considered optimal for initial anti-EGFR-based therapy supports both stop and go and maintenance as de-intensification strategies to consider in shared decision making. Adequately powered phase III studies are advocated to compare the two strategies. Introduction The first-line treatment of patients with microsatellite stable (MSS) metastatic colorectal cancer (mCRC) is a combination of fluoropyrimidine-based chemotherapy and anti-vascular endothelial growth factor (VEGF) or anti-epidermal growth factor receptor (EGFR) monoclonal antibodies, based on patient and tumor features, including primary tumor sidedness and molecular profile.1 Doublet chemotherapy plus an anti-EGFR agent is the preferred first-line treatment in patients with left-sided, RAS and BRAF wild-type (wt) disease.2, 3, 4, 5, 6, 7, 8, 9, 10 There is still a lack of evidence from phase III or adequately powered studies on the optimal maintenance or de-intensification strategy after this initial therapy.1,11,12 Several phase II studies have explored different schedules of maintenance therapy, such as anti-EGFR monotherapy, 5-fluorouracil/leucovorin (5-FU/LV), or their combination in RAS-wt tumors.13, 14, 15, 16 Additionally, intermittent treatment strategies (often referred to as ‘stop and go’) have been investigated.15, 16, 17 Based on the available evidence, after an induction with FOLFOX plus anti-EGFR, continuation maintenance with anti-EGFR plus 5-FU/LV is the guideline-recommended regimen. However, the magnitude of benefit from the addition of 5-FU/LV to anti-EGFR maintenance in patients with left-sided and BRAF-wt tumors is lower.12 In contrast, the optimal strategy after FOLFIRI plus anti-EGFR induction therapy is even less conclusive based on the available studies: continuation of the induction regimen until progression or unacceptable toxicity, anti-EGFR monotherapy maintenance, or an intermittent strategy has been investigated in this setting.15,17 Recently, the intermittent schedule showed a favorable safety profile and did not impair overall survival (OS) in patients with left-sided tumors.15,17 One of the main challenges in optimizing post-induction strategies is balancing efficacy with the management of toxicities. Anti-EGFR therapies are associated with significant adverse events (AEs), particularly skin toxicity, which may impact patients’ quality of life and adherence to treatment.18, 19, 20, 21, 22 Therefore, tailored strategies should be aimed at minimizing toxicity without compromising efficacy. Moreover, the potential OS impact of each strategy should be carefully evaluated, both overall and in specific patient subgroups. Considering the lack of adequately powered datasets, we designed and conducted an international individual patient data (IPD) pooled analysis including four randomized clinical trials (RCTs), aimed at comparing maintenance versus stop and go strategy after an anti-EGFR-based first-line treatment in patients with left-sided, non-microsatellite instability-high (MSI-H), RAS/BRAF-wt mCRC. Patients and methods Study design and trial patients We performed an international IPD pooled analysis of four multicenter randomized phase II clinical trials: PanaMa (NCT01991873), Valentino (NCT02476045), PRODIGE-28-TIME (NCT02404935), and IMPROVE (NCT04425239). Briefly, in the PanaMa trial, patients with RAS-wt mCRC who achieved disease control after six cycles of FOLFOX plus panitumumab induction were randomly assigned to maintenance therapy with 5-FU/LV alone or with panitumumab, followed by preplanned FOLFOX plus panitumumab reinduction after the first progression during maintenance. The trial showed significantly superior progression-free survival (PFS) but similar time to failure of strategy (TFS) with the combination maintenance.13,23 In the Valentino trial, after a 4-month induction with FOLFOX plus panitumumab in patients with RAS-wt mCRC, maintenance therapy with panitumumab monotherapy was inferior to the combination 5-FU/LV plus panitumumab in PFS, but postprogression therapy was not preplanned.14 The PRODIGE-28-TIME trial randomly assigned patients with RAS-wt mCRC after eight induction cycles of FOLFIRI plus cetuximab to maintenance therapy with single-agent cetuximab versus a chemotherapy-free interval, both followed by preplanned FOLFIRI plus cetuximab reinduction after progression. The primary endpoint of improvement in 6-month progression-free rate after randomization was not met, but PFS and OS favored cetuximab-based maintenance, especially for patients without activating mutations in mitogen-activated protein kinase (MAPK) pathway genes.15 In the IMPROVE trial, intermittent FOLFIRI plus panitumumab with a stop-and-go schedule was investigated as first-line treatment of patients with RAS/BRAF-wt mCRC as compared with continuous FOLFIRI plus panitumumab. The intermittent schedule was feasible, reduced severe skin toxicity, and was not associated with any detrimental effect on OS in patients with left-sided tumors.17 The eligibility criteria, detailed methods of the trials, along with their efficacy and safety results, have been previously published and presented in Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107775. This IPD pooled analysis included all patients who started the assigned post-induction therapy as per protocol. Only patients with RAS-wt, BRAF-wt (without the V600E mutation) and left-sided tumors were included in the analysis, while patients with known MSI-H status were excluded. Patients were stratified into three treatment groups based on their post-induction treatment regimen (maintenance with 5-FU/LV plus anti-EGFR, maintenance with anti-EGFR alone, or preplanned chemotherapy-free interval and reinduction after disease progression, defined as stop and go). The study protocols and their amendments were approved by the trial management committees of all trials and by an independent institutional review board or ethics committee. The study was conducted in compliance with Good Clinical Practice guidelines and the Declaration of Helsinki. All patients provided written informed consent. This IPD pooled analysis study was approved by the ethical committee of Fondazione IRCCS Istituto Nazionale dei Tumori and was promoted by Fondazione GONO Onlus (Identifier: INT 99/22). Data collection Data collection included anonymized collection of patients’ characteristics, that is, age, sex, Eastern Cooperative Oncology Group (ECOG) performance status, primary tumor location (left colon and rectum), primary tumor resection status, previous adjuvant treatment, onset of metastases (synchronous versus metachronous), number and site of metastases, first-line induction backbone, start and stop dates of induction and post-induction strategy, and best response to induction [stable disease (SD) versus partial or complete response (PR/CR)], number of cycles administered, treatment strategy (maintenance with 5-FU/LV plus anti-EGFR, anti-EGFR alone, or stop and go), treatment durations, and efficacy outcomes (disease progression and death), as previously described. A trial database was set up to include the information extrapolated from the four study datasets to ensure the collection of appropriately comparable data and to facilitate the planned IPD pooled analysis. Endpoints and statistical analysis The primary objective was to compare OS between a stop-and-go strategy and 5-FU/LV plus anti-EGFR maintenance. OS was defined as the interval from the start of the post-induction treatment strategy to death from any cause, with patients still alive censored at the last date of follow-up. Secondary objectives were to compare stop-and-go strategy with 5-FU/LV plus anti-EGFR maintenance in terms of PFS and TFS, and to compare stop-and-go strategy with anti-EGFR monotherapy maintenance in terms of OS, PFS, and TFS. PFS was defined as the time from the start of the post-induction treatment strategy to disease progression (PD) or death, whichever occurred first. TFS was defined as the time from initiation of the post-induction treatment strategy to PD during the treatment in patients who received stop and go or to second PD among patients who received maintenance followed by protocol-planned reinduction, provided reinduction was delivered; first PD if reinduction was not delivered after the first progression or if another treatment line (including at least a new drug nonadministered in the first induction) was initiated; start of another treatment line in case treatment strategy was withdrawn in absence of first PD; or death from any cause at any time, whichever occurred first. Patients who were alive without any of these events were censored at the date of last follow-up. As exploratory analyses, OS and TFS comparisons were also performed in the subgroup of patients receiving per-protocol reinduction, with outcomes measured from the start of post-induction treatment. Descriptive statistics were used to summarize the patients’ characteristics. Numerical variables were reported as medians with interquartile ranges (IQRs), while categorical variables were described using absolute and relative frequencies. The median follow-up time was calculated using the inverse Kaplan–Meier (KM) method. The KM curves of OS, PFS, and TFS were estimated stratifying by treatment strategy, and differences were assessed via the log-rank test. The impact of different post-induction strategies on the overall study cohort and subgroups was investigated using mixed-effects Cox proportional hazards models. As an IPD analysis was conducted on four distinct trials (Valentino, PanaMa, PRODIGE-28, and IMPROVE), adjustments for potential confounding factors and interstudy heterogeneity were necessary. For each comparison, a logistic regression model was constructed, with treatment strategy as the dependent variable and covariates including patient age, ECOG PS, primary tumor location, resection of primary tumor, prior adjuvant treatment, onset of metastases, number of metastatic sites, lung and liver metastases, and planned duration of induction. Inverse probability of treatment weighting (IPTW) was employed to generate the propensity score for each treatment strategy.24 A mixed-effects Cox regression model was then developed, incorporating treatment strategy as a fixed effect, a frailty term to account for trial heterogeneity, and propensity score as weighting factors.25 Propensity score was specifically derived for each comparison. Adjusted hazard ratios (HRs) along with 95% confidence intervals (CIs) were reported for each analysis. Subgroup analyses were performed comparing the effect of stop and go versus 5FU/LV plus anti-EGFR and versus anti-EGFR alone, considering age, sex, ECOG PS, primary tumor location, resection of primary tumor, prior adjuvant treatment, onset of metastases, number of metastatic sites, limited-liver metastases, and best response to the induction therapy as grouping variables. Each subgroup analysis was adjusted for confounders with IPTW. The Wald test for the interaction between each variable and treatment strategy group was performed. Statistical analyses were performed with R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). Results Study population The IPD included a total of 378 randomly assigned patients, specifically 134 in Valentino, 94 in Panama, 48 in Improve, and 102 in the PRODIGE-28 trial, respectively. Among 770 patients included in the initial pooled database, 392 were excluded as per prespecified criteria (they did not receive the planned post-induction treatment according to the three selected treatment groups or had a BRAF-mutated and/or MSI-H and/or right-sided tumor). Overall, 166 (44%), 109 (29%), and 103 (27%) patients received maintenance with 5-FU/LV plus anti-EGFR, maintenance with anti-EGFR alone, and stop-and-go strategy, respectively (Figure 1).Figure 1Flow chart of the IPD pooled analysis.5FU/FA, 5-fluorouracil leucovorin; cet, cetuximab; IPD, individual patient data; MSI, microsatellite instability high; pani, panitumumab; pts, patients; wt, wild type. Flow chart of the IPD pooled analysis. 5FU/FA, 5-fluorouracil leucovorin; cet, cetuximab; IPD, individual patient data; MSI, microsatellite instability high; pani, panitumumab; pts, patients; wt, wild type. Baseline patient and disease characteristics stratified by treatment group are illustrated in Table 1 and per individual trial in Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2026.107775. Briefly, baseline characteristics were well balanced in the three treatment groups except for ECOG PS, with the value of 2 being represented only in the stop and go and anti-EGFR alone groups. There was a lower proportion of rectum primary tumor site, one site of metastasis, and liver-limited disease in the stop and go group, and a higher proportion of synchronous metastases in the 5-FU/LV plus anti-EGFR maintenance group.Table 1Patients and disease characteristics according to treatment groupCharacteristicWhole seriesN = 378Stop and goN = 1035-FU/LV + anti-EGFR maintenanceN = 166Anti-EGFR alone maintenanceN = 109P valueAge, years65 (57-72)66 (59-72)65 (58-72)63 (54-69)0.019Sex0.296 Female116 (30.7)35 (34.0)44 (26.5)37 (33.9) Male262 (69.3)68 (66.0)122 (73.5)72 (66.1)ECOG PS0.007 0250 (66.8)67 (67.0)109 (65.7)74 (68.5) 1115 (30.7)26 (26.0)57 (34.3)32 (29.6) 29 (2.4)7 (7.0)02 (1.9) NA4301Primary tumor location<0.001 Left colon257 (68.0)90 (87.4)95 (57.2)72 (66.1) Rectum121 (32.0)13 (12.6)71 (42.8)37 (33.9)Primary tumor resected0.027 Yes227 (60.1)52 (50.5)111 (66.9)64 (58.7) No151 (39.9)51 (49.5)55 (33.1)45 (41.3)Prior adjuvant therapy<0.001 Yes64 (16.9)19 (18.4)24 (14.4)21 (19.3) No314 (83.1)84 (81.6)142 (85.5)88 (80.7) NA0000Onset of metastases0.010 Synchronous284 (75.1)69 (67.0)137 (82.5)78 (71.6) Metachronous94 (24.9)34 (33.0)29 (17.5)31 (28.4)Number of metastatic sites0.025 1202 (53.6)41 (40.2)96 (57.8)65 (59.6) 2127 (33.7)46 (45.1)48 (28.9)33 (30.3) >248 (12.7)15 (14.7)22 (13.3)11 (10.1) NA1100Liver-limited disease0.026 Yes144 (38.1)29 (28.2)74 (44.6)41 (37.6) No234 (61.9)74 (71.8)92 (55.4)68 (62.4)First-line backbone<0.001 Oxaliplatin-based228 (60.3)0166 (100)62 (56.9) Irinotecan-based150 (39.7)103 (100)047 (43.1)Best response to induction0.085 SD78 (20.6)29 (28.2)29 (17.5)20 (18.3) PR/CR300 (79.4)74 (71.8)137 (82.5)89 (81.7)Values are median (interquartile range) or n (%).5-FU, 5-fluorouracil; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, epidermal growth factor receptor; LV, leucovorin; maint., maintenance; NA, not available; PR, partial response; SD, stable disease. Table 1Patients and disease characteristics according to treatment groupCharacteristicWhole seriesN = 378Stop and goN = 1035-FU/LV + anti-EGFR maintenanceN = 166Anti-EGFR alone maintenanceN = 109P valueAge, years65 (57-72)66 (59-72)65 (58-72)63 (54-69)0.019Sex0.296 Female116 (30.7)35 (34.0)44 (26.5)37 (33.9) Male262 (69.3)68 (66.0)122 (73.5)72 (66.1)ECOG PS0.007 0250 (66.8)67 (67.0)109 (65.7)74 (68.5) 1115 (30.7)26 (26.0)57 (34.3)32 (29.6) 29 (2.4)7 (7.0)02 (1.9) NA4301Primary tumor location<0.001 Left colon257 (68.0)90 (87.4)95 (57.2)72 (66.1) Rectum121 (32.0)13 (12.6)71 (42.8)37 (33.9)Primary tumor resected0.027 Yes227 (60.1)52 (50.5)111 (66.9)64 (58.7) No151 (39.9)51 (49.5)55 (33.1)45 (41.3)Prior adjuvant therapy<0.001 Yes64 (16.9)19 (18.4)24 (14.4)21 (19.3) No314 (83.1)84 (81.6)142 (85.5)88 (80.7) NA0000Onset of metastases0.010 Synchronous284 (75.1)69 (67.0)137 (82.5)78 (71.6) Metachronous94 (24.9)34 (33.0)29 (17.5)31 (28.4)Number of metastatic sites0.025 1202 (53.6)41 (40.2)96 (57.8)65 (59.6) 2127 (33.7)46 (45.1)48 (28.9)33 (30.3) >248 (12.7)15 (14.7)22 (13.3)11 (10.1) NA1100Liver-limited disease0.026 Yes144 (38.1)29 (28.2)74 (44.6)41 (37.6) No234 (61.9)74 (71.8)92 (55.4)68 (62.4)First-line backbone<0.001 Oxaliplatin-based228 (60.3)0166 (100)62 (56.9) Irinotecan-based150 (39.7)103 (100)047 (43.1)Best response to induction0.085 SD78 (20.6)29 (28.2)29 (17.5)20 (18.3) PR/CR300 (79.4)74 (71.8)137 (82.5)89 (81.7)Values are median (interquartile range) or n (%).5-FU, 5-fluorouracil; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, epidermal growth factor receptor; LV, leucovorin; maint., maintenance; NA, not available; PR, partial response; SD, stable disease. Patients and disease characteristics according to treatment group Values are median (interquartile range) or n (%). 5-FU, 5-fluorouracil; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, epidermal growth factor receptor; LV, leucovorin; maint., maintenance; NA, not available; PR, partial response; SD, stable disease. Primary and secondary objectives OS and TFS curves stratified by trial are reported in Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2026.107775. Median follow-up was 65.3 (IQR 62.2-73.1), 62.4 (IQR 56.0-67.7), and 48.9 (IQR 47.0-56.9) months in 5-FU/LV plus anti-EGFR maintenance, anti-EGFR alone maintenance, and stop and go groups, respectively. In the analyses performed for the primary objective, no significant OS differences were reported. The median OS for stop and go and maintenance with 5-FU/LV plus anti-EGFR were 26.3 and 29.7 months (HR 1.24, 95% CI 0.65-2.37, P = 0.530 from the mixed-effects Cox model adjusted with propensity score), respectively (Figure 2A and Table 2).Figure 2Survival outcomes according to treatment group. Kaplan–Meier curves for OS (A), PFS (B), and TFS (C) calculated from the start of post-induction treatment according to treatment group comparing stop and go versus maintenance with anti-EGFR plus 5-FU/LV, respectively. Panels (D), (E), and (F) depict the Kaplan–Meier curves for OS, PFS, and TFS, respectively, which are calculated from the start of post-induction treatment according to treatment group, comparing stop and go vs maintenance with anti-EGFR alone.5-FU/LV, 5-fluorouracil/leucovorin; EGFR, epidermal growth factor receptor; OS, overall survival; PFS, progression-free survival; TFS, time to failure of strategy.Table 2Survival outcomes according to treatment strategyTreatmentOverall survivalProgression-free survivalTFSMedian (95% CI)HR (95% CI)P valueaMedian (95% CI)HR (95% CI)P valueaMedian (95% CI)HR (95% CI)P valueaStop and go26.3 (23.5-33.3)1.24 (0.65-2.37)0.5303.1 (2.5-4.1)2.78 (1.94-3.99)<0.000111.3 (7.8-13.2)1.04 (0.51-2.10)0.915anti-EGFR + 5-FU/LV29.7 (26.2-35.8)8.7 (7.6-10.4)10.8 (9.08-12.9)Stop and go26.3 (23.5-33.3)1.39 (0.87-2.22)0.1653.1 (2.5-4.1)1.85 (1.29-2.64)<0.000111.3 (7.8-13.2)1.14 (0.79-1.65)0.491Anti-EGFR alone30.1 (23.3-40.1)6.3 (5.3-8.9)10.7 (7.8-13.2)5-FU, 5-fluorouracil; CI, confidence interval; EGFR, epidermal growth factor receptor; HR, hazard ratio; LV, leucovorin; TFS, time to failure of strategy.aP values from mixed-effects Cox models adjusted with propensity score. Survival outcomes according to treatment group. Kaplan–Meier curves for OS (A), PFS (B), and TFS (C) calculated from the start of post-induction treatment according to treatment group comparing stop and go versus maintenance with anti-EGFR plus 5-FU/LV, respectively. Panels (D), (E), and (F) depict the Kaplan–Meier curves for OS, PFS, and TFS, respectively, which are calculated from the start of post-induction treatment according to treatment group, comparing stop and go vs maintenance with anti-EGFR alone. 5-FU/LV, 5-fluorouracil/leucovorin; EGFR, epidermal growth factor receptor; OS, overall survival; PFS, progression-free survival; TFS, time to failure of strategy. Table 2Survival outcomes according to treatment strategyTreatmentOverall survivalProgression-free survivalTFSMedian (95% CI)HR (95% CI)P valueaMedian (95% CI)HR (95% CI)P valueaMedian (95% CI)HR (95% CI)P valueaStop and go26.3 (23.5-33.3)1.24 (0.65-2.37)0.5303.1 (2.5-4.1)2.78 (1.94-3.99)<0.000111.3 (7.8-13.2)1.04 (0.51-2.10)0.915anti-EGFR + 5-FU/LV29.7 (26.2-35.8)8.7 (7.6-10.4)10.8 (9.08-12.9)Stop and go26.3 (23.5-33.3)1.39 (0.87-2.22)0.1653.1 (2.5-4.1)1.85 (1.29-2.64)<0.000111.3 (7.8-13.2)1.14 (0.79-1.65)0.491Anti-EGFR alone30.1 (23.3-40.1)6.3 (5.3-8.9)10.7 (7.8-13.2)5-FU, 5-fluorouracil; CI, confidence interval; EGFR, epidermal growth factor receptor; HR, hazard ratio; LV, leucovorin; TFS, time to failure of strategy.aP values from mixed-effects Cox models adjusted with propensity score. Survival outcomes according to treatment strategy 5-FU, 5-fluorouracil; CI, confidence interval; EGFR, epidermal growth factor receptor; HR, hazard ratio; LV, leucovorin; TFS, time to failure of strategy. P values from mixed-effects Cox models adjusted with propensity score. Regarding the secondary objectives, PFS was significantly shorter with stop and go compared with 5-FU/LV plus anti-EGFR maintenance, while TFS was similar between the two strategies; specifically, the median PFS was 3.1 versus 8.7 months (HR 2.78, 95% CI 1.94-3.99, P < 0.0001), and the median TFS was 11.3 versus 10.8 months (HR 1.04, 95% CI 0.51-2.10, P = 0.915), respectively (Figure 2B and C and Table 2). No significant differences were observed in OS with stop and go versus maintenance with anti-EGFR alone: median OS was 26.3 versus 30.1 months (HR 1.39, 95% CI 0.87-2.22, P = 0.165), respectively (Figure 2D and Table 2). PFS was significantly shorter with stop and go versus anti-EGFR alone, while TFS was comparable between the two strategies; specifically, the median PFS was 3.1 versus 6.3 months (HR 1.85, 95% CI 1.29-2.64, P < 0.0001), and median TFS was 11.3 versus 10.7 months (HR 1.14, 95% CI 0.79-1.65, P = 0.491), respectively (Figure 2E and F and Table 2). Subgroup analyses Across prespecified subgroups, no significant differences in OS, PFS, or TFS were observed between stop and go and 5-FU/LV plus anti-EGFR, with two exceptions. Among patients with ECOG PS ≥1, stop and go was associated with inferior OS versus 5-FU/LV plus anti-EGFR (HR 2.12, 95% CI 1.29-3.48, P-interaction = 0.009). For PFS, stop and go underperformed 5-FU/LV plus anti-EGFR in patients with ≥2 metastatic sites (HR 5.05, 95% CI 2.75-9.26 for patients with two metastatic sites and HR 4.10, 95% CI 1.82-9.23 for patients with >2 metastatic sites; P-interaction = 0.007). No significant interaction between post-induction treatment group and best response to induction was detected for OS, PFS, and TFS in the comparison of stop and go versus 5-FU/LV plus anti-EGFR. Conversely, when comparing stop and go versus anti-EGFR alone, PFS and TFS varied by best response to induction category. Stop and go was superior among patients with SD, whereas anti-EGFR alone was superior among patients achieving a PR/CR. Specifically, stop and go was also inferior as compared with anti-EGFR alone in patients achieving a PR/CR to induction, both in terms of PFS (HR 2.32, 95% CI 1.54-3.50, P-interaction = 0.007) and TFS (HR 1.33, 95% CI 0.88-2.03, P-interaction = 0.018; Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2026.107775). Notably, patients with PR/CR to induction showed a trend toward improved OS irrespective of the de-intensification strategy (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2026.107775). Exploratory analyses Among patients starting the protocol-defined post-induction strategy, the proportion of those receiving, as per protocol, at least one reinduction course after the first disease progression was 95% in the stop and go group, 34% in the 5-FU/LV plus anti-EGFR maintenance group, and 47% in the anti-EGFR alone group. Among these, OS and TFS were similar between stop and go and both maintenance strategies (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107775). Discussion In patients with mCRC, the choice of first-line treatment is a key point in the therapeutic algorithm, which should consider, on one hand, the therapeutic goal and the patient’s characteristics and preferences, and on the other, the tumor’s clinical and molecular profile.26 Major international guidelines identify patients with left-sided, MSS, RAS/BRAF-wt tumors as optimal candidates for an anti-EGFR-based first-line treatment.1 This is based upon the results of several pivotal studies that allowed refinement of the negative selection for anti-EGFR agents from the RAS-wt-only status to the identification of multiple clinical and molecular predictive markers of resistance, particularly tumor sidedness and BRAF V600E mutations.3,6,7,27,28 Concerning the balance between efficacy and toxicity, the major guidelines recommend treatment de-intensification after reaching disease control with a more intensive induction, especially one that is oxaliplatin-based.1 In the specific setting of an anti-EGFR-based induction, despite the lack of solid evidence derived from phase III or adequately powered RCTs, the current recommendations are based upon the results of several phase II trials addressing this relevant issue.13, 14, 15, 16, 17 Nevertheless, the cumulative toxicity of maintenance treatment, both skin-related for anti-EGFR agents and chemotherapy-induced, should be considered, given its potentially remarkable impact on quality of life.20, 21, 22,29 A crucial point to underline is that the most common reporting of AEs is the rate of occurrence of any-grade or grade 3 or higher AEs, though this underestimates the real burden of toxicity for the patient. The optimal reporting is a longitudinal evaluation or toxicity over time that allows for capturing the intensity and duration of singular AEs, considering even low-grade but long-lasting AEs that may affect quality of life more than short-lived, high-grade AEs. This is crucial, especially for skin-related toxicity from anti-EGFR treatment that is not relieved by a maintenance schedule that only de-potentiates the chemotherapy backbone, thus reducing chemo-related AEs but not anti-EGFR-related ones.30, 31, 32 A preplanned treatment break until the occurrence of PD, followed by reintroduction of the highly-active regimen, could significantly reduce the toxicity burden, as long as this would not reduce the treatment efficacy, especially in terms of long-term survival in adequately selected patients.15, 16, 17 Therefore, the optimal de-intensification strategy, maintenance or stop and go, after an anti-EGFR-based induction in mCRC patients selected to be the optimal candidates for this approach, is crucial. No direct comparison in ad hoc-designed clinical trials is available, given the limited feasibility of such a trial; thus, only indirect comparisons are possible, with all their intrinsic limitations. In this light, our effort to collect IPD from pivotal RCTs on this topic is relevant to derive clinically useful considerations. Indeed, it is the first, to our knowledge, comparison between maintenance and stop and go that addresses a daily clinical practice issue. Since chemotherapy continuation is particularly important in the subgroup of patients with more aggressive and treatment-resistant disease, such as right-sided tumors or BRAF-mutated ones, we decided to focus our comparisons on the subgroup of patients currently considered as the optimal candidates for anti-EGFR-based initial therapy, that is, those with left-sided, RAS/BRAF-wt mCRC. We decided to exclude patients with known MSI-H status, although MSI-H status was unknown for most patients, since they are treated with first-line immunotherapy and are relatively more resistant to anti-EGFR therapy than to bevacizumab.33 In our IPD pooled analysis, we reinforced the importance of the optimal definition of the first-line treatment, since we observed a trend toward improved OS in patients achieving response (PR/CR) to induction, irrespective of the type of de-intensification strategy received. Moreover, the interaction between treatment group and outcome did not provide meaningful results, because an advantage in PFS only, with no impact on OS, of stop and go versus anti-EGFR alone maintenance would not be considered reliable. Regarding the primary objective of the study, we did not observe significant differences in terms of OS between the stop and go schedule and maintenance. Therefore, the guideline-endorsed regimen of combination with anti-EGFR and fluoropyrimidines and the regimen of anti-EGFR monotherapy could both be options in clinically and molecularly selected patients.12 We should acknowledge that OS is also influenced by subsequent treatments after the end of the first-line strategy, which are not prespecified in clinical trials and are endowed with a considerable level of heterogeneity; however, OS is the most clinically relevant and standardized endpoint for this disease setting. Moreover, the absence of a significant difference in OS might be explained by the design and sample size of the study, in that it is possibly not adequately powered to detect an OS difference. Concerning the PFS results, in favor of maintenance, this is clearly expected according to the per-se design of the stop and go schedule, whereas TFS was considered a more appropriate endpoint as it includes the entire strategy of stop and go as well as maintenance and reinduction. In fact, comparable TFS results were reported for stop and go and maintenance in the per-protocol population with both maintenance schedules. Notably, reinduction was prespecified only in some RCTs included in the IPD analysis, and only a minor subgroup of patients truly received reinduction, thus possibly biasing the results. Nevertheless, the global results of TFS and OS are consistent, thus reinforcing our conclusions. In the subgroup analyses, no clear indications for a differential strategy in selected populations emerged; in fact, given the small sample size and the nature of the study, the results of the subgroup analyses should only be considered as hypothesis-generating. Therefore, our IPD analysis supports both strategies, stop and go and maintenance, as de-intensification treatments in patients with RAS/BRAF-wt, left-sided mCRC after an anti-EGFR-based induction. The choice of the therapeutic algorithm for individual patients should consider patient and tumor features, in particular ECOG PS and tumor burden, and treatment response and tolerability. In fact, in case of suboptimal ECOG PS due to disease and high burden, the preference could be for maintenance; conversely, in case of toxicity impacting quality of life, the stop-and-go schedule is naturally preferable. Another key point is the biological evolution of EGFR-addicted mCRC, since a well-consolidated burden of evidence has shown how the resistance mechanisms arise during first-line anti-EGFR-based therapy but may decrease during subsequent anti-EGFR-free treatment lines, creating the possibility for anti-EGFR rechallenge in further lines.34, 35, 36, 37 In the same light, the intermittent therapeutic approaches are supported by the rationale of reducing the selective pressure by anti-EGFR treatment on the tumor, possibly delaying the onset and spread of resistance.17 Interestingly, in the PanaMa study, maintenance with 5-FU/LV plus anti-EGFR improved PFS and conditioned a numerically longer OS as compared with 5-FU/LV alone, but was associated with reduced PFS of the protocol-planned reinduction and with a substantially comparable TFS.23 This could have been due to the gradual depletion of resistant clones in the anti-EGFR-free maintenance arm, possibly questioning the timepoint of combination of anti-EGFRs to the post-induction strategy. Nevertheless, the effort of clinicians is to always offer the optimal therapeutic schedule, even in the maintenance setting, since the risk of a rapid progression preventing the opportunity to continue the treatment strategy may be unavoidable. In addition, we are now able to better select patients from a clinical and molecular point of view, with left-sidedness as a good surrogate, thus allowing to obtain the maximal cytoreduction and to considerably reduce the risk of death or clinical deterioration at the first progression. Several clinical and translational studies clearly showed that the benefit from anti-EGFR therapy could be optimized by refining the molecular negative hyperselection beyond RAS and BRAF, such as including mutations of PTEN and EGFR extracellular domain exons 1-16, amplifications of HER2 and MET, and gene fusions of RET, NRTK1, and ALK. In tumors harboring other rarer and undetected molecular mechanisms of resistance, the benefit of maintenance treatment may be attenuated.5,9,38 The biological and exploratory analyses of previous and ongoing clinical trials on the stop-and-go strategy may provide relevant results on the topic of the optimal therapeutic management of this patient subgroup. The phase 3 IMPROVE2 (NCT06509126) trial is investigating first-line continuous or intermittent (stop and go) FOLFIRI plus panitumumab in unresectable left-sided RAS/BRAF-wt proficient mismatch repair and MSS mCRC. This trial will shed light on this topic and provide further evidence to help clinicians in the choice of the optimal therapeutic strategy for individual patients. Our study has several limitations, the first being the intrinsic design of the study as a pooled analysis of multiple clinical trials with different designs, regimens, and durations of induction, schedules of post-induction treatment, and follow-up duration. In particular, the issue of the intertrial heterogeneity in the choice of induction chemotherapy and the anti-EGFR agent selected is a relevant issue. Nevertheless, to account for heterogeneity, we performed an accurate statistical analysis with mixed-effects Cox models adjusted with propensity scores, including the most relevant confounding variables. However, the potential pitfall of using propensity scores, including the inability to adjust for unmeasured confounders, should be underlined. The small sample size in each treatment group is another limitation of the study. Moreover, knowing the impact on toxicity, in parallel with the efficacy results, would be fundamental to provide an integrated view of the effect of the treatment approach. Conclusions In optimal candidates for initial anti-EGFR-based therapy, our study supported both stop and go and maintenance as de-intensification strategies after an anti-EGFR-based induction, considering the comparable OS results. Shared decision making on the therapeutic algorithm should consider patient and tumor features, treatment response, and tolerability. Adequately powered phase III studies are advocated to compare the two strategies and provide conclusive evidence on this issue.