Real-World Data Analysis of Second-Line Antiangiogenic Targeted Treatments Following Anti-Epidermal Growth Factor Receptor Monoclonal Antibodies and First-Line FOLFOX for Patients with Metastatic Colorectal Cancer

Increased survival has been demonstrated in clinical trials for second-line use of bevacizumab [21], ramucirumab [13], and aflibercept beta [12, 22], and these antiangiogenic drugs are currently recommended in second-line therapy for patients with unresectable advanced or recurrent CRC [5, 7]. Consistent with these recommendations, our previous study, with a September 2019 cutoff, showed that these antiangiogenic drugs were used commonly as second-line therapy for patients with advanced CRC, regardless of first-line treatment choices [5, 6]. Despite this, information on the optimal treatment sequences for these patients, especially after first-line anti-EGFR therapy, is lacking. Each of the antiangiogenic drugs has specific molecular targets and mechanisms of action [23]; therefore, it is relevant to assess the differences in outcomes of therapy with these drugs. In this real-world study in Japanese patients with advanced CRC, median treatment duration from second-line chemotherapy with an antiangiogenic drug was 12.5 months, and there appeared to be no differences in treatment duration by antiangiogenic drugs. These findings are similar to previous reports that studied the effectiveness of antiangiogenic drugs after first-line anti-EGFR antibody therapy [24‐27]. However, results from retrospective studies need to be interpreted with caution and should be understood primarily as a sole description of real-world clinical practice because important clinical information for treatment decisions may not be available, and therefore comparative effectiveness evaluation of superiority among therapy choices with appropriate background adjustment is often difficult, especially in administrative database studies [28, 29].

The proportions of patients with a prescription for concomitant medications or procedures during second-line therapy were similar across each of the antiangiogenic drug groups, except for antihypertensive drugs, which were prescribed more frequently to patients receiving second-line ramucirumab and aflibercept beta than bevacizumab. This trend is similar to the higher incidence of hypertension with ramucirumab and aflibercept beta compared with bevacizumab in clinical trials [13, 22, 30], potentially leading to the differences in the frequency of antihypertensive drugs between the antiangiogenic treatment groups.

In the Cox regression analysis, we showed that duration of first-line therapy of 6 months or more, CRC surgery before starting first-line therapy, and liver surgery during first-line therapy were positively associated with treatment duration from second-line therapy. A retrospective study in Japan and a phase III RAISE trial of ramucirumab reported the prognostic role of progression-free survival (PFS) and time to progression, respectively, of first-line therapy for second-line irinotecan-based chemotherapy in patients with mCRC [31, 32]. As for CRC surgery before first-line therapy, an integrated analysis of eight randomized clinical trials reported that primary tumor resection for patients with synchronous mCRC was associated with better prognosis [33]. Also, the prognosis of patients with metachronous (recurrent) mCRC was better than for patients with unresected synchronous mCRC. These are in alignment with our results because our study population with CRC surgery before first-line therapy was likely to include patients with both resected synchronous and metachronous disease. However, a recent randomized clinical trial in Japan in patients with synchronous mCRC did not show benefit with primary tumor resection, and therefore the prognostic role of primary tumor resection is still under debate [34]. In the current study, 59.2% of patients had liver metastases and 5.9% had undergone liver resection during first-line therapy. The significant association of liver surgery with treatment duration is consistent with clinical practice (where approximately two-thirds of recurrent CRC involves the liver) and a study reporting that liver resection is associated with long-term survival [35]. However, the potential benefits of liver resection in this study need to be interpreted carefully because of the small sample size. In contrast, use of NSAIDs was a consistent negative factor in this study. This conflicts with a previous study reporting that appropriate use of NSAIDs improved the prognosis of patients with CRC [36]. A potential reason for this difference may be because the limited clinical information in our administrative database can lead to confounding (i.e., the use of NSAIDs for treatment of pain). Therefore, our results may suggest that conditions such as pain requiring NSAID treatment had an adverse impact on treatment continuation. ADL independence was another factor associated with treatment duration in our subgroup analysis, indicating the importance of the health status for treatment continuation. We also observed that patients’ health status, as indicated by comorbidities and ADL, deteriorated as patients progressed from before first-line to before third-line therapy. Overall, these results emphasize the importance of maintaining good health status among patients as they transition to subsequent lines of therapy.

Several other factors were identified in subgroup Kaplan–Meier analyses that may be associated with longer treatment durations. First, patients with left-sided tumors showed numerically longer treatment duration from second line than those with right-sided tumors. Primary tumor location is an established prognostic factor for treatment benefit in patients with CRC [7], and patients with left-sided RAS wild-type tumors in particular benefit from chemotherapy with anti-EGFR therapy [19, 20]. Also, a retrospective analysis of the FIRE-3 phase III trial reported numerically longer PFS and overall survival (OS) in patients with left-sided tumors for the second-line anti-vascular endothelial growth factor therapy after first-line cetuximab [37]. Second, treatment duration from second-line therapy was numerically longer in those who had used antihypertensives or proteinuria tests than in those who had not. As proteinuria and hypertension are known adverse events associated with the antiangiogenic drug class [38‐40], this suggests that patient management and monitoring for known adverse events during second-line therapy may contribute to improved outcomes. Furthermore, findings from a retrospective study showed that hypertension may be associated with the effectiveness of antiangiogenic therapies [41]. However, a causal relationship between prescription of antihypertensives or proteinuria tests and treatment duration cannot be confirmed because it is also possible that patients who remained in second-line therapy longer were more likely to have received an antihypertensive prescription or proteinuria test.

The current standard third-line therapy for patients with mCRC in Japan is limited to FTD/TPI monotherapy and the small-molecule drug, regorafenib [5]. In this study, treatment choices for third-line therapy were highly heterogenous. The durations of treatment for patients receiving third-line FTD/TPI monotherapy or regorafenib were numerically lower compared with third-line antiangiogenic- and anti-EGFR-based treatments. Patients receiving third-line FTD/TPI monotherapy or regorafenib appeared to be slightly older and have worse ADL status; therefore, it is possible that these therapies were chosen for more vulnerable patients, leading to shorter therapy continuation as a result. However, these findings are considered exploratory only because the availability of ADL data was limited to a small number of patients.

The main strength of this study is the use of a large-scale administrative claims database, which can provide evidence on the real-world use of prescription medication, monitoring procedures, and treatment patterns for patients receiving drug combinations that are not typically studied in clinical trials.

However, although administrative claims databases typically provide clinically relevant information on large numbers of patients, data are not collected for research purposes. Important clinical information on disease characteristics (e.g., cancer stage, performance status), effectiveness outcomes (e.g., OS, PFS, overall response rate), and reasons for discontinuation of therapies (e.g., death, tumor progression, toxicity) are not available or are incomplete in the MDV database, while others, such as some patient characteristics (e.g., ADL, BMI), can only be obtained from hospitalization records and may not be available for all patients. In addition, because diagnosis information in this database is from healthcare claims, it is possible that some claims records are left over from past diagnoses from which patients have recovered. Although information on RAS mutation status is not available in the MDV database, all patients included in this study were presumed to have RAS wild-type advanced CRC because anti-EGFR antibody therapy is only indicated for patients with this genotype [5, 42, 43]. In addition, because information that is essential for decision-making in clinical practice is incomplete in the MDV database, it is not feasible to adjust analyses for clinical characteristics or to conduct statistical comparisons of treatment outcomes between the treatment groups. Further, because the database only includes prescription information, the use of specific medications and care/monitoring concomitant therapies and procedures (e.g., proteinuria tests, antihypertensives) is inferred only, and treatment lines were defined using prespecified definitions of treatment sequences. Because a patient cannot be tracked between hospitals, patients may be lost or counted multiple times if they received treatment at more than one MDV hospital. Finally, any causal relationships between the identified factors and treatment outcome cannot be confirmed because of the small numbers of patients in some subgroups and because of the limitations of retrospective database analyses [44].