Real-world treatment patterns and effectiveness among patients with metastatic colorectal cancer treated with ziv-aflibercept in community oncology practices in the USA

Routine clinical practice data often differ from clinical trials. This study describes real-world treatment patterns and effectiveness among patients with metastatic colorectal cancer (mCRC) receiving ziv-aflibercept in non-academic, community oncology practices in the USA. De-identified electronic medical records from Vector Oncology and Altos Solutions databases were analysed. We identified 218 patients diagnosed with mCRC who had received prior oxaliplatin therapy and initiated ziv-aflibercept as part of second-line or later-line therapy. Overall survival (OS) and progression-free survival (PFS) were estimated using Kaplan–Meier analysis. Mean age was 62.8 years at ziv-aflibercept initiation. Most patients (91.7%) received bevacizumab before ziv-aflibercept, 95.4% initiated ziv-aflibercept with FOLFIRI or another irinotecan-based regimen, and 59.6% had received prior irinotecan. Overall, 24.8% of patients initiated ziv-aflibercept in second line, 31.7% in third line, 21.6% in fourth line and 22.0% in later lines of therapy. Mean duration of ziv-aflibercept treatment was 5.3 months. For patients initiating ziv-aflibercept in second-, third- and fourth-line therapy, median OS was 11.9 (95% confidence interval 5.1–16.2), 11.1 (6.9–16.7) and 8.1 (5.2–11.4) months, respectively, and median PFS was 4.4 (2.8–6.5), 4.3 (2.9–6.3) and 3.4 (2.2–5.2) months, respectively. Common adverse events (AEs) (any grade) included gastrointestinal disorders (64.7%) and asthenia/fatigue (63.3%). In routine clinical practice, ziv-aflibercept was frequently initiated in third line or later lines of therapy. Although patients receiving ziv-aflibercept were more heavily pretreated and potentially less robust compared with the VELOUR trial, median OS for patients receiving second-line ziv-aflibercept was comparable. AE rates were similar to or lower than the VELOUR trial.

193 Page 2 of 10 studies have evaluated vascular endothelial growth factor (VEGF)-targeted agents such as bevacizumab, ziv-aflibercept and ramucirumab in combination with chemotherapy and have reported statistically significant improvements in overall survival (OS) for patients with mCRC. A phase 3 trial demonstrated an OS and progression-free survival (PFS) benefit with bevacizumab when added to irinotecan, fluorouracil and leucovorin (IFL) compared with IFL plus placebo in previously untreated patients with mCRC [5]. In addition, the phase 3 'RAISE' trial demonstrated an OS benefit with ramucirumab plus FOLFIRI compared with FOLFIRI plus placebo in patients with mCRC with disease progression during or after first-line therapy with bevacizumab, oxaliplatin and a fluoropyrimidine [6].
The randomized, double-blind, phase 3 'VELOUR' trial evaluated the efficacy and safety of ziv-aflibercept plus FOL-FIRI as second-line treatment for patients with mCRC who had disease progression on or after oxaliplatin-based treatment [7]. Prior use of bevacizumab was permitted, but all patients were irinotecan-naïve. Compared with placebo plus FOLFIRI, the addition of ziv-aflibercept to FOLFIRI significantly improved both OS (13.5 vs. 12.1 months; hazard ratio [HR] 0.82; p = 0.0032) and PFS (6.9 vs. 4.7 months; HR 0.76; p < 0.0001). Anti-VEGF treatment-related toxicities such as hypertension, mucosal bleeding and proteinuria were reported in more patients in the ziv-aflibercept arm compared with the control arm.
As clinical trials often include a select patient population, it remains uncertain whether the results of a new agent in a clinical trial can be replicated in routine 'real-world' clinical practice. To date, no studies have reported treatment patterns and effectiveness in real-world clinical practice among patients with mCRC treated with ziv-aflibercept. The objectives of this study were to assess real-world characteristics, treatment patterns, effectiveness and adverse events (AEs) in patients with mCRC who have initiated ziv-aflibercept treatment as second-line or later-line therapy following a prior oxaliplatin-based regimen.

Study design
This retrospective cohort study aimed to assess patients with mCRC included in the Vector Oncology Data Warehouse and Altos Solutions Inc. Database who had initiated ziv-aflibercept as second-line or later-line therapy following an oxaliplatin-based treatment. The databases are comprehensive cancer patient databases derived from networks of community oncology practices in the USA, comprising demographic information, medical history, treatment information, laboratory results and clinical information from electronic medical records (EMRs) and billing data. This study included review of de-identified structured EMR data (from Jan 2003 to Jul 2014) and a manual review of unstructured preexisting EMR data (such as physician notes and EMR attachments) for fields that are not well populated in the structured EMR dataset [Health Insurance Portability and Accountability Act (HIPAA) compliant data were collected]. The study protocol was reviewed and approved by the New England Independent Review Board. Because of the retrospective nature of this study, patient informed consent was not required in accordance with institutional regulations.

Inclusion criteria
Eligible patients had mCRC and had received ziv-aflibercept (second line or later) not as part of a clinical trial. Patients were required to have received oxaliplatin therapy prior to ziv-aflibercept initiation. In addition, patients were required to have no other primary tumours during the baseline period (1-year period before ziv-aflibercept initiation).

Outcomes measures
The date of ziv-aflibercept initiation was defined as the index date. The follow-up period started at the date of zivaflibercept initiation and ended at the data cut-off date, last clinic visit date or death. Baseline clinical characteristics were assessed before or at the index date of ziv-aflibercept initiation, and the distribution of the line of ziv-aflibercept initiation was described. Lines of therapy after mCRC diagnosis were identified using the following algorithm based on chemotherapy/targeted therapy administrations. The start of first-line therapy was defined as the first administration of chemotherapy or targeted therapy after diagnosis of mCRC. The first-line regimen contained all chemotherapy-related drugs administered within 28 days of the start of the regimen. Subsequent lines of therapy were identified by the first administration of any chemotherapy-related drug not used in the previous line of therapy. The subsequent regimens contained all chemotherapy-related drugs that were administered within 28 days of the first administration in that line of therapy. If there was a gap of ≥ 120 days between treatment administrations within a given line of therapy, the second treatment was considered a new line of therapy. Chemotherapy or targeted therapy administrations were identified based on review of generic names of administered medications, review of generic names for oral medication prescriptions and information for oral chemotherapy medications. The end date for each line of therapy was defined as the last date of administration of any of the drugs used in the regimen before the start of the subsequent line of therapy. Duration of ziv-aflibercept treatment was measured in days from the first administration of ziv-aflibercept to the last Page 3 of 10 193 administration of ziv-aflibercept. Duration of the chemotherapy regimen containing ziv-aflibercept was measured in days from the first administration of the first drug administered in the regimen to the last administration of any of the chemotherapy drugs used in the regimen before the start of the subsequent line of therapy. Patients were considered censored if the last administration of either the ziv-aflibercept or the concomitant chemotherapy regimen was within 60 days of the date of last contact. Reasons behind discontinuation of ziv-aflibercept and the concomitant regimen, reasons for treatment interruptions and dose changes, and the date of last contact/death were also collected.
Other outcomes data (effectiveness, all treatment-emergent AEs and resource utilization) were also assessed. Objective tumour response was based on information available in the EMR notes; to the extent information was available, tumour response was further classified as complete or partial. OS was defined as the time from ziv-aflibercept initiation to death or last contact (for censored patients). PFS was defined as the time from ziv-aflibercept initiation to first disease progression or death (whichever occurred earliest) or last contact (for censored patients). Information about treatment-emergent AEs during ziv-aflibercept treatment was extracted for prespecified events, including those reported in the ziv-aflibercept prescribing information. For resource utilization, the number of patients with hospitalizations and emergency room visits after ziv-aflibercept initiation was reported.

Statistical analyses
Statistical analysis was performed on the Altos Solutions and Vector Oncology datasets, both individually and on the pooled data. Data extracted from the individual databases are reported for baseline patient and clinical characteristics in this publication; remaining data and analyses are shown for the pooled dataset.
To assess the heterogeneity of the populations from the two data sources, meta-analysis random effects models were estimated for select baseline characteristics, including age, gender, race, tumour location, number of sites of metastases, occurrence of liver metastases, presence of hypertension, months from first mCRC diagnosis to ziv-aflibercept initiation, and prior bevacizumab use [8].
Continuous variables were summarized by mean, median, standard deviation and interquartile ranges (or minimum and maximum). Categorical variables were summarized by absolute frequencies and percentages. Kaplan-Meier analyses were performed to evaluate time to discontinuation of the entire chemotherapy regimen concomitant with ziv-aflibercept, time to ziv-aflibercept discontinuation, PFS and OS. Survival rates at 6 and 12 months were also reported. These variables were analysed both in the overall population and according to the line of therapy at which ziv-aflibercept was initiated. All data manipulations and statistical analyses were performed using SAS version 9.3 software.

Baseline patient and disease characteristics
In total, 218 patients were included in the final study population (Vector Oncology database, n = 107; Altos Solutions database, n = 111). The mean age at ziv-aflibercept initiation was 62.8 years ( Table 1). The majority of patients were male (61.0%) and Caucasian (59.6%). Notably, the majority of the patient population (93.5%) identified from the Vector Oncology database were geographically located in the South of the USA, while the patient population identified from the Altos Solutions database was more evenly distributed among the Northeast (30.6%), South (30.6%) and West (23.4%). Heterogeneity tests found no significant (p < 0.05) heterogeneity between populations for most baseline characteristics except months from mCRC diagnosis to ziv-aflibercept initiation and prior bevacizumab use.
In patients with available biomarker data for KRAS (n = 179), 46.4% had wildtype KRAS and 53.6% had a KRAS gene mutation; among patients with biomarker data for BRAF (n = 46), 82.6% had wildtype BRAF and 17.4% had a BRAF gene mutation. The most common site of metastasis was the liver (79.4%). The most frequent conditions diagnosed prior to ziv-aflibercept initiation were hypertension (55.5%) and tobacco use (29.8%).

Treatment patterns
Overall, 24.8% of patients initiated ziv-aflibercept as secondline therapy for mCRC ( Table 1). The majority of patients initiated ziv-aflibercept treatment in the third-line or laterline setting, with 31.7% of patients receiving ziv-aflibercept as third-line therapy, 21.6% as fourth-line therapy, 11.9% as fifth-line therapy and 10.1% of patients initiating zivaflibercept beyond the fifth-line setting. The median initial dose of ziv-aflibercept in 2-week treatment cycles was 4 mg/ kg. Dose adjustments of ziv-aflibercept occurred in 9.2% of patients, mainly resulting from AEs/toxicity (80.0%). The majority of patients received bevacizumab (91.7%) or irinotecan (59.6%) as part of therapy for mCRC and prior to ziv-aflibercept initiation. In total, 90.4% of patients initiated ziv-aflibercept with FOLFIRI, of which 48.7% had previously received FOLFIRI. Among patients initiating ziv-aflibercept with irinotecan (95.4%), 59.6% of patients had previously received irinotecan. After initiation of zivaflibercept, the median length of follow-up was 7.1 months. The median duration of treatment with ziv-aflibercept was 193 Page 4 of 10  Fig. 1a). The median duration of treatment with the first chemotherapy regimen including ziv-aflibercept was 4.5 months (Table 1 and Fig. 1b). The median duration of ziv-aflibercept treatment was slightly longer among patients who initiated ziv-aflibercept in second-line therapy (4.9 months; 95% confidence interval [CI] 2.1-6.0) or in third-line therapy (5.1 months; 95% CI 2.8-7.0) relative to the overall population. The most frequent reasons for discontinuation of ziv-aflibercept were disease progression (49.0%) and AEs/toxicity (23.4%).

Discussion
Randomized clinical trials are essential for establishing the efficacy and safety of new drugs, and for informing regulatory approval and clinical practice algorithms. Nonetheless, clinical trials often enrol a narrowly defined patient population who are treated under highly controlled parameters. As such, studies that examine the use, effectiveness and safety of interventions in patients treated outside of clinical trials, in the context of routine clinical practice, can further inform the practical utility of a new agent. This retrospective analysis studied patients with mCRC who received ziv-aflibercept as second-line or later-line therapy following a prior oxaliplatin-based treatment regimen. In this sample of USA patients treated with ziv-aflibercept outside of a clinical trial and within non-academic community oncology practices, treatment characteristics differed from patients in the phase 3 trial of ziv-aflibercept plus FOLFIRI as second-line treatment in patients with mCRC (VELOUR). In the VELOUR study, ziv-aflibercept was administered as part of second-line treatment in all patients. In contrast, in this real-world analysis only 24.8% of patients initiated zivaflibercept as second-line therapy, more than two-thirds of patients initiated ziv-aflibercept as third-line or later-line therapy, and six patients (2.9%) initiated ziv-aflibercept in the seventh line or later. Also, nearly 60% of patients in this analysis received prior irinotecan-based therapy compared with the VELOUR trial, which required patients to be irinotecan-naïve. In addition, over 90% of patients in this analysis received bevacizumab prior to the initiation of zivaflibercept, compared with only 30% of patients enrolled in the VELOUR trial [7].
Despite patients in this analysis being more heavily pretreated-and potentially less physically fit-compared with the VELOUR trial, the median OS for patients receiving second-line ziv-aflibercept was approximately 12 months for both studies [7]. The median OS was also comparable to other previously published trials of second-line anti-VEGF therapy in patients with mCRC, including the ECOG Study E3200 (FOLFOX4 plus bevacizumab in mCRC; median OS 11.9 months) and the RAISE trial (ramucirumab plus FOLFIRI in mCRC; 13.3 months) [6,9]. In this analysis, approximately 20% of all patients responded to treatment. Almost 26% of patients responded to second-line treatment, which is comparable with both the VELOUR (19.8%) and ECOG E3200 (22.7%) studies [6,9].
The rate of treatment-emergent AEs experienced by patients in this analysis was similar to or lower than the