Materials and methods
Study setting and eligibility criteria of studies
This systematic review was performed following the recommendations of the Cochrane Handbook for Systematic Reviews (25) , Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) (26) , and AMSTAR 2 guidelines (27) . We previously registered the protocol in Prospective Register of Systematic Review (PROSPERO) (CDR 42022364692, registered on October 14, 2022).
We included studies that evaluated adults diagnosed with non-small cell lung cancer (NSCLC) by histological or cytological methods and who had a mutation in the epidermal growth factor receptor (EGFR). The participants were required to have a functional status score of 2 or lower, according to the criteria of the Eastern Cooperative Oncology Group (ECOG). Our review focused on randomized clinical trials (RCTs) that compared the combination of erlotinib and bevacizumab versus the use of erlotinib alone in the treatment of NSCLC with positive EGFR mutation. No restrictions were set in terms of race, gender, nationality, histological type, or smoking history of the participants. Studies that were reviews without primary data or extractable data, animal experiments or cadaver studies, patients with other types of cancer, patients who received previous treatment, and meta-analyses were excluded.
Database and search strategy
We searched for clinical trial randomized studies published until April 15, 2023 in Medline (PubMed), Scopus, and Embase. We combined different keywords, controlled vocabulary terms (e.g., MeSH and Emtree terms), and free terms, according to the PICO strategy (population: “carcinoma, non-small-cell lung”; exposure: “erlotinib hydrochloride” AND “bevacizumab”; comparator: “erlotinib hydrochloride”) (Supplementary Material). The searches were not limited by date or language. We included articles in full text and excluded observational studies, review articles, abstracts, case reports, letters, editorials, studies not available in full text, and duplicated publications.
Study selection and data extraction
We exported all retrieved references from databases to Rayyan QCRI (Rayyan Systems Inc.®, MA, USA). After removing duplicates, two authors (VEFR and ATM) performed independently the screening of title and abstracts. These authors independently reviewed the remaining references in the full text. Discrepancies were resolved by a third researcher. The references from retrieved papers were screened for additional articles. The articles found were analyzed using the terms of the PICO strategy and the inclusion and exclusion criteria. Relevant data from each article were extracted by two authors (VEFR and MJVG) independently and recorded in a spreadsheet of Microsoft Excel©: name of authors, year and country of publication, number of patients, number of events, and measure of association, with their 95% confidence intervals. Any conflict regarding the extracted information was resolved through consensus.
Quality assessment
Two investigators (VEFR and ATM) independently evaluated the risk of bias in each eligible RCT. Any discrepancies were resolved by consensus or discussion with another investigator (MJVG). The Cochrane Collaboration tool for assessing the risk of bias in RCTs was used (28) . The following items were evaluated: generation of the allocation sequence (selection bias), concealment of the allocation sequence (selection bias), blinding (detection and performance bias), blinding of participants and personnel to outcome assessment, incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), and other biases; each domain is a question and is answered according to yes/no/not known to rate it as high risk/low risk/unknowledge risk. Each RCT underwent a meticulous evaluation, with outcomes for each item categorized as having either a low risk of bias, a high risk of bias, or an unclear risk of bias.
Outcome measures
Primary outcome variables were progression-free survival (PFS) defined as the time from randomization to tumor progression or death. Secondary outcomes were overall survival (OS) defined as the time from randomization to death, considered as the best therapeutic endpoint in cancer clinical trials, overall response ratio (ORR) defined as the proportion of patients whose symptoms were relieved to a predetermined value within the minimum time limit, and serious adverse event (SAE) defined as adverse event of grade 3 or more.
Statistical analyses
In the meta-analysis, we pooled hazard ratios (HR) with 95% confidence intervals (95% CI) using fixed-effects models and followed the inverse variance method (due to the large number of events in each arm, more than 10%). The Paule–Mandel estimator was used for the assessment of the between-study variance (29) . Outcomes data available in ≥3 studies were meta-analyzed. Time-event variables, including OS and PFS, were assessed according to the HR. Dichotomous variables, including ORR and incidence of adverse events, were assessed as risk ratios (RR) with 95% confidence interval (CI) estimates. For studies reporting OR or RR stratified into different subgroups, we considered each subgroup analysis as a separate study. The quantitative synthesis was represented by forest plots. Heterogeneity among studies was assessed with Cochran’s Q test and Higgins I2 statistics; values of I2 <20%, 25%–50%, and >50% were defined as low, mild, and substantial heterogeneity, respectively. If I2 was <50% and p > 0.05, a fixed-effects model was used in the meta-analysis; if I2 ≥50% and p ≤ 0.05, a random-effects model was used to assess the resource of the heterogeneity. Publication bias was assessed with funnel plots and formally tested with Egger’s test.
All statistical analyses were conducted using RevMan version 5.4, provided by the Cochrane Collaboration. A p-value less than 0.05 was considered to indicate statistical significance.
Discussion
This study provides new insights that could help resolve the controversies surrounding the combined use of erlotinib and bevacizumab in the treatment of EGFR-mutated NSCLC.
The EGFR gene encodes the protein located on the cell surface whose activation stimulates the molecular pathways that allow the growth and development of the tumor microenvironment (3, 5) . As previously described, osimertinib is the preferred TKI for the first line of treatment of EGFRm metastatic NSCLC (11, 12) . Several trials evaluated different combinations that could be safe and effective in this population, and the combination of angiogenesis inhibitors and TKI obtained discordant results (21–24) . Currently, osimertinib is the agent of choice for first-line treatment due to its greater penetrance in the CNS. However, the economic cost of osimertinib limits its access to clinical practice. An economic analysis reported that osimertinib as the first line of treatment is not cost-effective in high-income countries (32–34). Subsequent studies have evaluated possible combinations that could be options of therapy (3).
Our results show a statistically significant benefit in terms of progression-free survival (PFS) (HR = 0.63; 95% CI, 0.54–0.73). This result is consistent with those obtained in clinical trials and is similar to those obtained by other recently published meta-analyses (35–38) . The studies JO25567, ARTEMIS, NEJ026, and Stichcombe et al. obtained positive results in progression-free survival when evaluating the addition of an angiogenic inhibitor (bevacizumab) to the TKI (erlotinib) compared to a TKI given as a monodrug in first-line treatment for advanced NSCLC with EGFRm (9) . In 2014, one of the first clinical trials that was phase II (JO25567) showed that the addition of bevacizumab to erlotinib in patients with NSCLC improved PFS from 9.7 to 16.0 months in Japanese patients, with a HR of 0.54 (0.36 to 0.81). Another phase II clinical study (Stichcombe 2019) was the only study that did not report a statistical benefit of bevacizumab on PFS (HR: 0.81; 95% CI: 0.50–1.31); however, this study found a clinical benefit in mPFS of 17.9 vs. 13.5 months for the group of bevacizumab with erlotinib vs. erlotinib alone, respectively, similar to the phase II study by Lee et al. (2023) (31), which found no statistical benefit but observed a higher median PFS months (17.5 vs 12.4 months).
Another statistically significant result was ORR which was higher with the bevacizumab–erlotinib combination (RR = 0.79; 95% CI, 0.64–0.97, p = 0.03). However, higher ORR has not been replicated in all meta-analysis, probably because of heterogeneity (31, 35–40) . Our analysis only included clinical trials where the combination of bevacizumab and erlotinib was used to maintain a homogeneous population (I2 = 0%, p = 0.66).
Finally, no statistically significant benefit in overall survival (OS) was demonstrated (HR = 0.93; 95% CI, 0.78–1.10). It should be noted that the bevacizumab–erlotinib combination did not improve OS in clinical trials (21–24) . The consistent lack of benefit in OS could be explained by the subsequent line of treatment. Patients with progression of disease received osimertinib as second line when T790M mutation was detected in blood or tumoral tissue. Patients were treated with osimertinib after progression of disease: 29.2%, 57.1%, and 43% of patients in the erlotinib group and 17.2%, 49%, and 45% of patients in the bevacizumab–erlotinib group in the ARTEMIS, BEVERLY, and NEJ026 trials, respectively (21, 30, 41) . Based on the details above, we believe that there could be a methodological limitation that affects the accuracy of the OS results; this could affect the accuracy of OS results due to the difference in the intensity of subsequent therapy between the two groups.
SAEs were more common with bevacizumab–erlotinib combination. However, despite the higher toxicity, clinical trials conclude that bevacizumab–erlotinib is safe, with manageable toxicity (21–23) . Another meta-analysis indicates that the angiogenesis inhibitors and TKI combination is safe in NSCLC patients (38, 42) . Using the combination in first-line therapy may lead to the sequency of treatment with osimertinib in second-line. Clinical trials report a similar prevalence of T790M mutation after first-generation TKIs and the combination progression of disease. In addition, an economic analysis reported that osimertinib is cost-effective in second-line (43, 44).
In a subgroup analysis, the results indicate that the bevacizumab–erlotinib combination is associated with longer PFS in exon 19 deletion or exon 21 L858R mutations, suggesting that this combination may be similarly effective in both molecular subtypes. This result is consistent with that obtained in a recently published meta-analysis, suggesting that L858R mutations may be as much benefited as exon 19 deletion (38) . Further investigation is needed to conclude if this combination of treatment is the best option in L858R mutations.
The presence of brain metastasis is a recognized adverse prognostic factor and a frequent site of disease progression in EGFR-mutated NSCLC. Previous research has shown that the anti-VEGF medication bevacizumab is a successful treatment for brain metastases and can also significantly decrease their incidence (45).
Bevacizumab has demonstrated a significant clinical benefit in CNS tumors, highlighting its anti-edema effect. This effect is particularly relevant in the reduction of post-radiation brain edema, a common complication in patients with brain metastases (46, 47); in addition, bevacizumab has shown a beneficial effect in the treatment of refractory brain edema with an efficacy rate of 84.74% (46). Bevacizumab improves vascular permeability and reduces fluid infiltration into brain tissue, which helps mitigate the adverse effects of edema and improves treatment efficacy (47). This mechanism may explain the improvement in progression-free survival observed in patients with brain metastases following a combination treatment with erlotinib and bevacizumab.
Grade ≥3 AEs were more commonly reported with bevacizumab–erlotinib combination (diarrhea, rash, hypertension, and proteinuria). The grade ≥3 toxicity described is mainly secondary to bevacizumab. The toxicity of the combination therapy should be considered according to the patient’s comorbidities when making treatment choices. Patients with chronic hypertension that is not controlled or chronic kidney disease may not be candidates for the bevacizumab–erlotinib combination. In addition, the implementation of a strict monitoring program of blood pressure measurement and urinalysis in patients receiving bevacizumab must be done when treating these patients.
Recent studies have highlighted significant advances in predicting treatment response in advanced NSCLC. On the one hand, research has shown that changes in tumor density and volume measured by CBCT during therapy can correlate with clinical outcomes, offering a promising method for predicting treatment response at an early stage. On the other hand, radiomic signatures based on pre-treatment CT images have demonstrated their ability to predict PD-L1 expression and TMB status, and these signatures can be combined with clinical and morphological factors to improve the predictive efficacy and targeting of immunotherapy (48, 49).
Recently conducted clinical trials on osimertinib and bevacizumab have failed to demonstrate any improvement in either PFS or OS as compared to osimertinib alone among patients suffering from NSCLC with an EGFR mutation, regardless of whether they are undergoing first-line (24) or second-line treatment (50). Despite the existence of only phase II studies, the potential for combination therapy appears to be unpromising.
The enhanced PFS value observed when utilizing erlotinib-plus-bevacizumab may be attributed to modifications in tumor vessel physiology caused by bevacizumab, ultimately leading to greater intratumoral drug uptake and improved drug delivery (51).
As we have indicated above, without considering aspects of economic studies, third-generation TKIs (osimertinib) are considered as first-line treatment in some regions of the world based on the analysis of multiple measures for second-line treatment after drug resistance. However, we believe that some patients may miss the therapeutic opportunity to receive treatment with third-generation TKIs due to failure to detect the presence of the T790M mutation. The FLAURA trial showed that first-line osimertinib is similar to second-line osimertinib and still has a very beneficial effect (17), but more evidence is needed to determine the most effective and reasonable treatment line in the use of osimertinib. Accordingly, we believe that the combination of erlotinib + bevacizumab may be very useful in those countries where osimertinib is not cost-effective (32–34) or in patients where the T790M mutation has not been detected (or it is not possible to sequence).
A recently published systematic review and meta-analysis by Sakharkar et al. (2023) (52) compared the efficacy and safety of the combination of erlotinib and bevacizumab with erlotinib monotherapy in patients with advanced first- or second-line NSCLC. The combination of erlotinib plus bevacizumab significantly prolonged PFS (HR: 0.62, 95% CI: 0.56–0.70, p < 0.001) but did not show a significant improvement in OS (HR: 0.95, 95% CI: 0.83–1.07, p = 0.39) and ORR (HR: 1.03, 95% CI: 1.02–1.07, p = 0.01) and ORR (HR: 1.10; 95% CI: 0.98–1.24, p = 0.09). Our results report a similar HR in PFS, but slightly lower in OS, and our results (evaluated in the first-line setting) found a benefit of the combination in ORR (RR: 0.79; p = 0.003). A network meta-analysis (53) aimed at comparing different first-line regimens for NSCLC treatment in an Asian population found that the highest scoring regimens were erlotinib (Erl) plus Bev (SUCRA: 0.94) and Erl plus ramucirumab (Ram) (SUCRA: 0.93) (but only one study) compared to bevacizumab alone (SUCRA: 0.87). These results, which are like those found in our study, suggest that while there are benefits in terms of PFS, the combination may carry additional risks that should be considered but can be considered as a good therapeutic option. We found similar hazard ratios (HRs) for both exon 19 deletions and exon 21 L858R mutations. This suggests that the combination therapy may be effective across different EGFR mutation subtypes, offering a potentially valuable treatment option for patients with these common EGFR mutations. Given the distinct molecular pathways activated by different EGFR mutations, our findings provide a basis for tailoring treatment approaches based on specific mutation profiles.
Based on the details above, we believe that, in clinical practice, a patient who is unable to use osimertinib and must instead receive erlotinib may benefit from the combination with bevacizumab, given the presence of exon 19 or exon 21 mutation, brain metastasis, and no contraindication of anti-angiogenic. Regarding future research directions, it is essential to conduct further studies to evaluate the effectiveness and safety of different combinations and sequences of TKIs in advanced NSCLC. Ongoing clinical trials (NCT02856893 and NCT03790397) are examining the optimal sequence of gefitinib and osimertinib in first-line treatment, which may provide insight into the best treatment strategy. This combination could be particularly beneficial for patients with brain metastases, given bevacizumab’s demonstrated effectiveness in treating CNS tumors and reducing post-radiation brain edema. However, it is crucial to consider individual patient factors such as comorbidities and potential for adverse events when selecting this combination therapy. Our study highlights the need for accessible and effective treatment options in regions with limited resources and underscores the importance of expanding global access to a broader range of therapies for NSCLC.
The present meta-analysis has limitations that should be considered. Firstly, the quantitative synthesis comprised six studies, some with little sample sizes, phases of clinical trials, and follow-up periods as previously stated. This heterogeneity may affect the results. Secondly, the meta-analysis was conducted at the trial level rather than the individual patient level. A sensitivity analysis was infeasible due to a limited number of eligible studies, many of which were open-label. Therefore, potential prognostic factors, patient comorbidities, extent of disease, and other genetic mutations were not examined in our study, potentially constraining our analyses. Despite the study’s limitations, the rigorous process of selecting studies enabled a comprehensive evaluation of the available evidence related to the topic of interest. We adhered to the PRISMA, Cochrane, and PROSPERO guidelines.
In conclusion, the study shows that the bevacizumab–erlotinib combination significantly improves PFS and ORR in EGFRm metastatic NSCLC. Bevacizumab–erlotinib is also associated with higher grade ≥3 adverse events. Although toxicity is manageable, a patient’s comorbidities must be strongly considered when using treatment with the bevacizumab–erlotinib combination. In addition, the combination may be an option in the first-line in countries without access to osimertinib. Prospective trials are needed to validate the benefit in L585R mutations.
Author contributions
RM-G: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. ALG-L: Conceptualization, Writing – original draft, Writing – review & editing, Supervision, Validation. VF-R: Conceptualization, Writing – original draft, Writing – review & editing, Data curation, Formal Analysis, Methodology. AC-V: Conceptualization, Validation, Writing – original draft, Writing – review & editing. RV-C: Conceptualization, Methodology, Writing – original draft, Writing – review & editing. YH-C: Investigation, Visualization, Writing – original draft, Writing – review & editing. AT: Conceptualization, Data curation, Formal Analysis, Writing – original draft, Writing – review & editing. MV-G: Conceptualization, Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing. VR-F: Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing. CC: Conceptualization, Supervision, Validation, Writing – original draft, Writing – review & editing. LR: Supervision, Validation, Writing – original draft, Writing – review & editing.