Introduction

In the therapeutic management of patients with RAS and BRAF wild-type (WT) metastatic colorectal cancer (mCRC), chemotherapy in combination with monoclonal antibodies directed against vascular endothelial growth factor (VEGF) or epithelial growth factor receptor (EGFR) are standard options in the first-line setting [1]. Previous studies have highlighted the impact of left-sided primary tumor location to identify patients who benefit from the selection for anti-EGFR therapy combination [2,3,4]. Moreover, negative hyperselection of patients by the absence of gene alterations associated with anti-EGFR resistance (i.e. mutations in BRAF V600E, KRAS, NRAS, PIK3CA, and PTEN, amplifications in HER2 and MET, and fusions in ALK, RET, and NTRK1) can optimize treatment stratification [5,6,7,8,9].

De-intensification of first-line treatment with oxaliplatin-based doublet or triplet therapies using maintenance strategies demonstrated improved tolerability and quality of life without compromising survival outcomes, compared to the continuation of intensive induction therapy until disease progression or treatment holidays [10,11,12,13]. After anti-EGFR based first-line treatment, several maintenance strategies were previously investigated, ranging from single-agent anti-EGFR antibody, fluoropyrimidine monotherapy or the combination of 5-fluorouracil/leucovorin (5-FU/LV) with an anti-EGFR antibody [14,15,16,17,18,19,20].

Based on the evidence from multiple phase 2 randomized clinical trials, the current European Society of Medical Oncology (ESMO) consensus guideline recommends maintenance therapy with 5-FU/LV plus an anti-EGFR after FOLFOX (5-FU/LV and oxaliplatin) plus anti-EGFR induction [1, 16, 18, 21]. In a previous pooled analysis of four randomized clinical trials investigating anti-EGFR based maintenance strategies, the combination of 5-FU/LV plus an anti-EGFR was confirmed as the optimal treatment strategy in this setting, especially in patients with stable disease to induction therapy, right-sided primary tumors or BRAF mutated status [21]. However, the combined impact of primary tumor sidedness and hyperselection status on the efficacy of maintenance strategies such as anti-EGFR or 5-FU/LV mono- or combination-therapy remains unclear. Given the highest predicted benefit of additional anti-EGFR based first-line in patients with left-sided primary tumors and absence of alterations associated with resistance to anti-EGFR treatment, the magnitude of the benefit of 5-FU/LV added to anti-EGFR maintenance in this subgroup is uncertain according to the risk/benefit ratio.

In this international individual patient data (IPD) pooled analysis of the randomized PanaMa (NCT01991873) and Valentino (NCT02476045) phase 2 trials that investigated anti-EGFR or 5-FU/LV mono- or combination maintenance regimens, we aimed to understand the individual and combined prognostic value of primary tumor sidedness and hyperselection status, and their predictive impact on the efficacy outcomes of the individual maintenance strategies, especially in the left-sided, negative hyperselected subgroup.

Results

Patient characteristics

A total of 607 (PanaMa: 378; Valentino: 229) patients received induction therapy. Of these, valid information on primary tumor sidedness and hyperselection status was available for 589 (PanaMa 360; Valentino: 229) and 511 (PanaMa: 303; Valentino: 208), respectively (Fig. 1, Induction set). Of these, 487 (80.2%) and 388 (63.9%) comprised the left-sided and negative hyperselected induction set, respectively (Fig. 1, Table 1).

Fig. 1: Patients flow diagram of the study population.
figure 1

Populations during induction and maintenance therapy of the PanaMa and Valentino trials and the respective subsets with available data on primary tumor sidedness and hyperselection status.

Full size image
Table 1 Baseline characteristics of all patients included in the analysis, and stratified by primary tumor location and hyperselection status (induction set).
Full size table

A total of 412 (PanaMa: 248; Valentino: 164) patients received maintenance therapy. Of these, valid information on primary tumor sidedness and hyperselection status was available for 401 (PanaMa: 237; Valentino: 164) and 360 (PanaMa: 202; Valentino: 158), respectively (Fig. 1, Maintenance set). Of these, 336 (83.8%) and 287 (79.7%) comprised the left-sided and negative hyperselected maintenance set, respectively (Fig. 1, Supplementary Table 1).

Baseline characteristics of subgroups by primary tumor sidedness (left- and right-sided) and hyperselection status (negative hyperselected and gene altered) revealed significantly increased fractions of female patients in the right-sided subgroup (Chi-square P < 0.001; Table 1); liver-limited disease in the left-sided, and the negative hyperselected subgroups, respectively (Chi-square P = 0.025 and Chi-square P = 0.029; Table 1). Remaining baseline characteristics were well comparable across all subgroups (Table 1).

Prevalence of genomic alterations

In 511 patients with available hyperselection status, a total of 136 genomic alterations were detected across 123 (24.1%) patients. Mutations were detected with a prevalence of BRAF V600E, N = 42 (34.2%); PIK3CA, N = 31 (25.2%); RAS, N = 29 (23.6%); PTEN, N = 14 (11.4%); AKT1, N = 5 (4.1%), ERBB2, N = 1 (0.8%). Overexpression/amplification of HER2 was detected in 14 patients (11.4%; Fig. 2; Supplementary Table 2). In 12 (9.8%) patients, more than 1 alteration was detected, of which 1 (0.8%) patient, had 3 detectable alterations (Fig. 2; Supplementary Table 2).

Fig. 2: Oncoprint diagram of gene alterations (mutations, overexpressions/amplifications) detected in the induction set.
figure 2

Each column represents one patient with at least one mutation. detected. Black bars in the top panel represent the cumulative prevalence of alterations per patient. The prevalence of the respective alteration is displayed on right. The bottom panel represents corresponding clinical information.

Full size image

Prognostic impact of primary tumor sidedness and hyperselection status on efficacy of induction therapy

Response rates to induction treatment were significantly higher in patients with left-sided primary tumors compared to those with right-sided primaries (69.2% vs. 49.0%; Chi-square P < 0.001; Table 1). Significantly higher response rates to induction treatment were also observed in patients in the negative hyperselected group compared to those in the gene altered group (69.1% vs. 55.3%; Chi-square P = 0.005; Table 1).

In patients receiving induction therapy, both left-sided primary tumor versus right-sided and negative hyperselected status versus gene altered status were associated with a significantly prolonged induction PFS (HR: 0.74; 95% CI: 0.59–0.93; P = 0.009 and HR: 0.65; 95% CI: 0.50–0.85; P < 0.001, respectively; Fig. 3a, b) and induction OS (HR: 0.60; 95% CI: 0.47–0.76; P < 0.001 and HR: 0.57; 95% CI: 0.43–0.75; P < 0.001, respectively; Fig. 3c, d). Combining sidedness with hyperselection, the best outcomes in terms of induction PFS and induction OS were observed in the left-sided, negative hyperselected subgroup, compared to worst outcomes in the right-sided, gene altered subgroup. Patients in the left-sided, negative hyperselected subgroup, compared to the right-sided, gene altered subgroup, significant benefits were noted in terms of induction PFS (HR: 0.53; 95% CI: 0.38–0.74; P < 0.001; Fig. 3e) and induction OS (HR: 0.37; 95% CI: 0.26–0.51; P < 0.001; Fig. 3f). In the left-sided, negative hyperselected subgroup, compared to the left-sided, gene altered, there was a significant benefit in induction PFS and induction OS (HR: 0.63; 95% CI: 0.48–0.83; P < 0.001 and HR: 0.62; 95% CI: 0.47–0.83; P < 0.001, respectively; Fig. 3e, f) and a numerical benefit in terms of induction PFS and induction OS compared to the right-sided, negative hyperselected subgroups (HR: 0.85; 95% CI: 0.62 - 1.17; P = 0.297 and HR: 0.82; 95% CI: 0.58–1.15 P = 0.255, respectively; Fig. 3e, f). In the right-sided population, patients in the negative hyperselected subgroup had numerically improved induction PFS (HR: 0.81; 95% CI: 0.65–1.00; P = 0.054; Fig. 3e) and significantly prolonged induction OS (HR: 0.70; 95% CI: 0.56–0.89; P = 0.002; Fig. 3f), compared to the gene altered subgroup.

Fig. 3: Progression-free and overall survival of induction therapy according to primary tumor sidedness, hyperselection status and the combination of both.
figure 3

a Progression-free survival, b overall survival stratified by primary tumor sidedness. c Progression-free survival, d overall survival stratified by hyperselection status. e Progression-free survival, f overall survival stratified by combination of primary tumor sidedness and hyperselection status.

Full size image

Restricting the analysis to alterations with established impact on anti-EGFR efficacy routinely tested in clinical practice (RAS, BRAF V600E, HER2) maintained the prognostic significance of negative hyperselection and primary tumor sidedness, aligning with the findings obtained using the full alteration panel (Supplementary Table 3).

Impact of primary tumor sidedness and hyperselection status on the efficacy of maintenance strategies

In patients receiving maintenance therapy, a left-sided primary tumor was associated with a significant benefit in maintenance PFS in patients receiving 5-FU/LV plus Panitumumab (HR: 0.72; 95% CI: 0.56–0.93; P = 0.010; Fig. 4a) or Panitumumab alone (HR: 0.68; 95% CI: 0.49–0.95; P = 0.049; Fig. 4a) when compared to 5-FU/LV monotherapy, while no significant benefit in maintenance OS was noted (Fig. 4b). In patients with left-sided primary tumors, the addition of 5-FU/LV to Panitumumab maintenance was not associated with a significant benefit in maintenance PFS or maintenance OS (HR: 1.06; 95% CI: 0.78–1.43; P = 0.680 and HR: 1.11; 95% CI: 0.80–1.54; P = 0.584, respectively; Fig. 4a, b). In patients with right-sided primary tumors, maintenance PFS and maintenance OS were comparable irrespective of maintenance treatment, except for a significant maintenance OS benefit for the addition of Panitumumab to 5-FU/LV (HR: 0.59; 95% CI: 0.27–0.94; P = 0.029; Supplementary Fig. 1a, b).

Fig. 4: Progression-free and overall survival of maintenance therapy according to maintenance treatment arms in the left-sided, negative hyperselected, and the combined left-sided; negative hyperselected subgroups.
figure 4

a Progression-free survival, b overall survival of patients with left-sided primary tumors. c Progression-free survival, d overall survival of patients with negative hyperselected status. e Progression-free survival, f overall survival of patients with combination of left-sided primary tumors and negative hyperselected status.

Full size image

The negative hyperselected subgroup showed a significant improvement in maintenance PFS in patients receiving 5-FU/LV plus Panitumumab (HR: 0.68; 95% CI: 0.51–0.91; P = 0.009; Fig. 4c) or Panitumumab alone (HR: 0.63; 95% CI: 0.44–0.91; P = 0.028; Fig. 4c) when compared to 5-FU/LV monotherapy, while no significant benefit in maintenance OS was noted (Fig. 4d). In patients with negative hyperselected tumors, the addition of 5-FU/LV to Panitumumab maintenance was not associated with a significant benefit in maintenance PFS or maintenance OS (HR: 1,08; 95% CI: 0.78–1.50; P = 0.604 and HR: 0.89; 95% CI: 0.63–0.126; P = 0.486, respectively; Fig. 4c, d). In patients in the gene altered subgroup, maintenance PFS and maintenance OS were comparable irrespective of maintenance treatment (Supplementary Fig. 2a, b).

In the left-sided, negative hyperselected subgroup, there was a significant improvement in maintenance PFS in patients receiving 5-FU/LV plus Panitumumab (HR: 0.67; 95% CI: 0.49–0.90; P = 0.010; Fig. 4e) or Panitumumab alone (HR: 0.58; 95% CI: 0.39–0.85; P = 0.010; Fig. 4f) when compared to 5-FU/LV monotherapy, while no significant benefit in maintenance OS was noted (Fig. 4f). However, in the left-sided, negative hyperselected subgroup, the addition of 5-FU/LV to Panitumumab maintenance was not associated with a benefit in maintenance PFS (median PFS, Panitumumab monotherapy: 9.4 months; 95% CI: 7.5–11.3 and median PFS, 5-FU/LV plus Panitumumab: 8.8 months; 95% CI: 7.1–10.5; HR: 1.16; 95% CI: 0.82–1.64; P = 0.371; Fig. 4e) or maintenance OS (median OS, Panitumumab monotherapy: 39.0 months; 95% CI: 24.0–56.0 and median OS, 5-FU/LV plus Panitumumab: 33.7 months; 95% CI: 27.0–40.5; HR: 1.06; 95% CI: 0.73–1.54; P = 0.795; Fig. 4f), irrespective of the inclusion of BRAF V600E into the gene altered panel (Supplementary Table 4).

In the right-sided subgroup, stratification of hyperselection status by BRAF V600E status, restriction to alterations with established impact on anti-EGFR efficacy routinely tested in clinical practice (RAS, BRAF V600E, HER2), and including all alterations revealed significantly improved PFS and OS in patients without BRAF V600E receiving 5-FU/LV plus Panitumumab compared to those with BRAF V600E (Supplementary Table 5). As the remaining hyperselection definitions showed only a numerical trend, these findings underscore the predictive impact of BRAF V600E status in this subgroup.

Restricting the analysis to alterations with established impact on anti-EGFR efficacy routinely tested in clinical practice (RAS, BRAF V600E, HER2) maintained the significant impact of negative hyperselection and primary tumor sidedness, aligning with the findings obtained using the full alteration panel (Supplementary Table 6).

In patients receiving 5-FU/LV plus Panitumumab as maintenance therapy, the pooled analysis demonstrated that hyperselection status and primary tumor sidedness significantly impacted treatment outcomes. In contrast, analyses of the individual trial cohorts did not reach statistical significance for impact of these factors on maintenance treatment outcomes (Supplementary Table 7).

Discussion

This IPD pooled analysis of the PanaMa and Valentino trials provides intriguing insights into the prognostic and predictive roles of primary tumor sidedness and hyperselection status in patients with RAS WT mCRC receiving an anti-EGFR and oxaliplatin-based first-line treatment followed by maintenance strategy. Results of this analysis confirm the prognostic impact of these stratification factors during induction therapy in terms of PFS and OS and outline their potential to predict maintenance treatment efficacy in terms of PFS, especially in the left-sided, negative hyperselected subgroup.

Our results verify that among patients receiving anti-EGFR and FOLFOX-based first-line treatment, those with left-sided primary tumors and negative hyperselected status have a more favorable prognosis compared to those with right-sided, gene-altered tumors. These findings align with previous studies highlighting the importance of tumor location and hyperselection status in predicting response to anti-EGFR based first-line therapies [2,3,4,5,6, 8, 22, 23]. Primary tumor sidedness is indicated in guidelines as a selection factor for the first-line treatment choice and its prognostic value was confirmed in this IPD pooled analysis [1]. Additional molecular selection beyond sidedness is fundamental to optimize RAS wild-type mCRC patient stratification in the maintenance setting, as previously reported [7, 9]. The combination of both factors underlined their prognostic potential and confirmed the left-sided, negative hyperselected as the best and right-sided, gene altered as the worst subgroup in terms of PFS and OS outcomes for an anti-EGFR-based first-line strategy. This was underlined by the enrichment of patients with negative hyperselected status in patients in the maintenance set as compared to the induction set.

However, it is worth noting that in a subset of patients in both trials RAS and BRAF mutations were identified during the hyperselection analysis. While RAS mutations were exclusion criteria, the secondary identification of RAS mutations in both trial cohorts underlines the significance of central and sensitive RAS testing to potentially identify subclonal drivers of early resistance. In addition, patients with BRAF-mutated mCRC are now commonly excluded from anti-EGFR therapy but were included in both trial populations [1, 16, 18]. However, our observations in the left-sided, negative hyperselected subgroup were validated irrespective of BRAF-status. To assess the impact of RAS, RAF and HER2 alterations, a revised hyperselection definition, that included exclusively alterations routinely assessed for anti-EGFR resistance (RAS, BRAF V600E, HER2) was used to re-analyze the impact on treatment outcomes. This restricted analysis yielded prognostic findings consistent with those obtained using the broader gene panel, underscoring the robustness of hyperselection status as a biomarker in induction and maintenance settings and underscoring the impact of broadly assessed alterations in guiding maintenance therapy in this setting.

During maintenance therapy, both anti-EGFR based treatment arms showed significant improvement in PFS compared to 5-FU/LV monotherapy both in the individually left-sided or negative hyperselected subgroups, and in the combination of both subgroups. However, no statistically significant or apparently clinically meaningful PFS benefit for adding 5-FU/LV to anti-EGFR based could be observed in any of these subgroups. In line, previous individual analyses of sidedness and hyperselection status revealed their predictive impact on anti-EGFR based maintenance strategies [7, 9]. In PanaMa, addition of Panitumumab to 5-FU/LV maintenance resulted in significantly prolonged OS in patients with right-sided primary tumors, despite limited numbers in these subgroups [18]. In agreement with our results, a combined analysis of primary tumor sidedness and negative hyperselection from the PARADIGM trial described a numeric OS benefit for the right-sided, negative hyperselected cohort for a Panitumumab vs. Bevacizumab based first-line in mCRC [24]. Additionally, previous studies have demonstrated Amphiregulin expression as a predictive biomarker of anti-EGFR efficacy, also in the right sided colon, as a potential explanation of this observation [25, 26]. In line with the previous individual trial reports, no significant benefit in OS was observed across any of the subgroups, suggesting that while maintenance strategies can delay disease progression, although the existing trials were not powered to detect differences in OS [16, 18].

Our study has several limitations, including the retrospective nature of the IPD pooled analysis and potential heterogeneity between the included trials and the fact that other maintenance studies such as MACRO-2, or COIN-B, or studies with FOLFIRI-based induction such as ERMES and PRODIGE-28 were not included [14, 15, 27, 28]. In addition, the methods and definitions for negative hyperselection were slightly divergent in the PanaMa and Valentino trials, i.e. the use of HER2 in situ hybridization (ISH) was restricted to the Valentino cohort, while in PanaMa only HER2 3+ immunohistochemistry (IHC) and ERBB2 mutations were considered to assess HER2 alterations [7, 9]. Patients who did not have disease progression during induction and reached the maintenance phase may be positively selected, potentially not representing the overall population, particularly respective to RAS/BRAF mutations and right-sided primary tumors, hampering conclusions to be drawn from these subpopulations. Finally, evaluation of combined primary tumor sidedness and hyperselection status on efficacy of maintenance strategies was limited to the left-sided, negative hyperselected population due to limited numbers in the remaining subgroups. The comparator arms from PanaMa and Valentino were analyzed separately, and no formal direct comparison was intended. As such, any contextual interpretation of outcomes across these groups should be made with caution, given the methodological limitations of cross-trial comparisons. However, pooled patient cohorts enhanced statistical power, and while separate subgroup analyses corroborated our findings, the combined analysis provided a more robust assessment of the biomarkers’ predictive value. In routine practice, patients with left-sided tumors that are RAS and BRAF WT, HER2-negative, and microsatellite stable are considered optimal candidates for anti-EGFR based treatment strategies. Although this study additionally explored more extensive molecular hyperselection parameters, the present data do not establish a clear added value for these markers beyond the current standard-of-care. Despite these limitations, our findings provide evidence for novel maintenance treatment stratification approaches in mCRC.

In clinical practice, these results support the stratification of patients based on primary tumor sidedness and hyperselection status when considering maintenance therapy options following oxaliplatin- and anti-EGFR based induction. Stratification of patients with left-sided, negative hyperselected tumors to anti-EGFR based maintenance could optimize treatment choices in this population, potentially sparing patients from additional toxicity, clinic time, and financial burden associated with 5-FU/LV, helping to increase the overall quality of life in this patient population without compromising PFS outcomes. However, the skin toxicity related to anti-EGFR treatment should be considered also in light of the potential use of stop and go strategies, especially in adequately selected patients based on depth of response to therapy, disease burden, extended molecular profile and patients’ preferences in a shared decision making process [29,30,31]. Further investigation into the relative contribution of individual hyperselection parameters in this setting is warranted.

In conclusion, our analysis highlights the significant prognostic value of primary tumor sidedness and hyperselection status in mCRC. Personalized maintenance strategies based on these factors can enhance treatment efficacy and improve patient outcomes.

Methods

Study design and patients

This was a secondary, exploratory analysis of a previously published IPD pooled analysis including the PanaMa and Valentino ranomized phase 2 trials [16, 18, 21]. In the Panama trial (NCT01991873), maintenance therapy involving Panitumumab combined with 5-FU/LV demonstrated significantly superior PFS compared to 5-FU/LV alone, in patients with RAS WT mCRC who achieved at least disease control after six cycles of Panitumumab plus FOLFOX. The Valentino phase II trial (NCT02476045) employed a randomization to a 4-month induction phase with panitumumab plus FOLFOX, followed by maintenance therapy with Panitumumab alone or the same induction treatment followed by maintenance with Panitumumab plus 5-FU/LV, in patients with RAS WT mCRC. The trial aimed at demonstrating non-inferiority of maintenance therapy with single-agent Panitumumab as compared to its combination with 5-FU/LV in terms of 10-month PFS. Non-inferiority was not proved and the results showed that Panitumumab was inferior in terms of PFS compared to panitumumab plus 5-FU/LV, although the combination arm had a slightly increased treatment-related toxicity.

The eligibility criteria, detailed methods of the studies, along with efficacy and safety results, have been reported previously [16, 18]. The present analysis included all patients who received the pre-planned induction therapy for analyses regarding induction therapy and all patients who started the assigned maintenance regimen for analyses regarding maintenance therapy. The study protocols and its amendments were approved by an independent institutional review board or ethics committee at each study site. The study was conducted in compliance with Good Clinical Practice guidelines and the Declaration of Helsinki. All patients provided written, informed consent. This secondary analysis is part of an IPD pooled analysis study and was approved by the ethical committee of Fondazione IRCCS Istituto Nazionale dei Tumori, which assessed consent to and granted authorization for future research use and data sharing (Identifier: INT 99/22) [21].

Data collection

Anonymized data collection included patients’ characteristics, i.e. age, sex, Eastern cooperative oncology group (ECOG) performance status, primary tumor location, hyperselection status (negative hyperselected vs. gene altered), primary tumor resection status, previous adjuvant treatment, number and site of metastases, first-line treatment start and stop dates, number of cycles administered and best response to induction therapy, treatment arm (maintenance with Panitumumab, maintenance with 5-FU/LV monotherapy, maintenance with Panitumumab plus 5-FU/LV) treatment durations, and efficacy outcomes (disease progression and survival), as previously described [21]. A trial database was set up to include the information extrapolated from the two study datasets to ensure the collection of appropriately comparable data and to facilitate the planned IPD pooled analysis. Primary tumor sidedness was assessed as left-sided (splenic flexure to rectum) versus right-sided (caecum to transverse colon) primary tumors.

Molecular analyses for negative hyperselection were previously reported for PanaMa and Valentino, respectively [7, 9]. Briefly, in PanaMa, negative hyperselection was performed using targeted next-generation sequencing (NGS; Illumina Cancer Hotspot Panel v2) to detect mutations in RAS, BRAF (V600E only), PIK3CA, AKT1, ERBB2, PTEN and IHC for HER2 overexpression/amplification [9]. In addition, in Valentino, IHC and dual-color silver ISH for HER2 and MET were performed. Actionable gene fusions were screened using IHC for ALK/ROS1/panTRK/RET; in all samples with evidence of IHC staining of any intensity/extension, whole-transcriptome shotgun sequencing was performed to confirm the presence of specific rearrangements. Oncogenic mutations in RAS, BRAF (V600E only), PIK3CA, AKT1, ERBB2, PTEN were assessed using NGS (Illumina Cancer Hotspot Panel v2) [7]. While presence of RAS mutations precluded inclusion in the original trials, they used assays with a limited RAS mutation panel, subsequent testing with more comprehensive panels identified additional mutations and they were included in this analysis.

For this IPD pooled analysis, selection criteria for both trials were combined to achieve comparable definitions for hyperselection status. Namely, patients were negatively hyperselected by the presence of mutations in RAS, BRAF (V600E only), PIK3CA, AKT1, ERBB2, PTEN or HER2 overexpression detected by IHC and ISH as previously described [7, 9]. Of note, for patients from the PanaMa cohort, no ISH was available, so only patients with a HER2 IHC score 3+ were defined as HER2-positive. Patients presenting with at least one of the aforementioned alterations were grouped in the gene altered cohort. If no alteration was detected, patients were defined as negative hyperselected.

Endpoints and statistical analysis

Patients were grouped and compared based on the specific maintenance treatment received: 5-FU/LV + anti-EGFR, anti-EGFR monotherapy, or 5-FU/LV alone. The primary outcome measures were progression-free survival (PFS) and overall survival (OS) in the overall study population and within key subgroups, including primary tumor sidedness and hyperselection status. PFS was defined as the time from the start of induction (induction PFS) or maintenance (maintenance PFS) treatment to disease progression (PD) according to RECIST 1.1 criteria or death from any cause, while OS was defined as the time from the start of induction (induction OS) or maintenance (maintenance OS) treatment to death from any cause. If patients did not experience progression or death, data were censored at the last date known event-free. Two sets of patients with available information on sidedness and hyperselection status were defined for the prognostic and predictive analyses, respectively. The induction set of patients included all patients who received at least one cycle of first-line induction therapy and was used for the prognostic analyses. The maintenance set of patients included all patients who received at least one cycle of maintenance therapy and was used for the predictive analyses.

Clinical and biological characteristics of patients according to the maintenance treatment arms were summarized using standard descriptive statistics and compared using Pearson’s Chi-squared test. PFS and OS curves were estimated using the Kaplan–Meier method. The impact of different maintenance regimens on the overall study cohort and various subgroups was investigated using Cox proportional hazards regression models.

Exploratory analyses were performed on the main subgroups based on the following baseline characteristics: sex (male vs. female), age (>65 vs. ≤65 years), ECOG PS (0 vs. 1), primary tumor resection (yes vs. no), prior adjuvant therapy (yes vs. no), synchronous vs. metachronous presentation of metastases, number of metastatic sites (1 vs. >1), primary tumor sidedness (left- vs. right-sided), hyperselection status (negative hyperselected vs. gene altered), sites of metastases (liver-limited vs. non-liver-limited), and overall response rate to first-line treatment according to RECIST 1.1 criteria (progressive disease [PD] or stable disease [SD] vs. complete response [CR] or partial response [PR]). Given the similar eligibility criteria, treatment protocols and endpoint definitions in both trials, and supported by prior IPD analyses, the maintenance chemo plus Panitumumab arms were pooled to increase statistical power and precision in evaluating this ESMO-recommended treatment strategy. As the comparator arms (chemo alone from PanaMa and panitumumab alone from Valentino) were not pooled, direct comparisons between these groups should be interpreted with caution due to the limitations of cross-trial analyses.

Results from the models were summarized using hazard ratios (HR) with corresponding 95% confidence intervals (CI) and likelihood ratio test P values. P values < 0.05 (two-sided) were considered statistically significant. SPSS PASW 29.0 (SPSS, Chicago, IL) and R v4.3.2 (The R Foundation for Statistical Computing, Vienna, Austria) software were used for statistical analyses.