Introduction

Systemic therapy for metastatic colorectal cancer (mCRC) generally begins with chemotherapy regimens involving a fluoropyrimidine in combination with either oxaliplatin or irinotecan [1]. Biologic agents are often added to these cytotoxic backbones. Specifically, chemotherapy may be combined with monoclonal antibodies targeting the vascular endothelial growth factor pathway, such as bevacizumab, or the epidermal growth factor receptor, namely cetuximab or panitumumab; this latter class of antibodies is only effective in patients with RAS and BRAF-wild-type tumors [2].

Two oral agents have been studied and approved for mCRC that has progressed on the above therapies [2]. Regorafenib is an oral multikinase inhibitor approved by the U.S. Food and Drug Administration in 2012 [3] based on the results of the phase III CORRECT trial, comparing regorafenib with placebo in patients with previously treated mCRC who had progressed beyond their last available standard therapy [4]. The trial demonstrated a significant improvement in overall survival (OS) among patients receiving regorafenib compared with those receiving placebo (median OS, 6.4 vs. 5.0 months; hazard ratio [HR], 0.77; p value = .005) [4]. Trifluridine/tipiracil (FTD/TPI) is an oral chemotherapy agent consisting of a cytotoxic component, trifluridine, and a thymidine phosphorylase inhibitor, tipiracil. FTD/TPI was approved in 2015 for the same population of patients for whom regorafenib was approved [5]; this approval was based on the results of the phase III RECOURSE trial, which compared FTD/TPI with placebo in patients with mCRC who had received at least two prior standard chemotherapy regimens [6]. This study showed a significant improvement in OS among patients treated with FTD/TPI compared with placebo (median OS, 7.1 vs. 5.3 months; HR, 0.68; p value <.001) [6].

However, real-world evidence of clinical outcomes for patients treated with regorafenib and FTD/TPI is lacking. This information is important given regorafenib and FTD/TPI have equivalent indications and routes of administration. This retrospective comparative effectiveness study used data collected from electronic medical records (EMRs) and a clinical database at a tertiary oncology center to evaluate and compare the impact of FTD/TPI versus regorafenib on clinical outcomes such as tumor response and survival among patients with mCRC who had progressed on or were intolerant to previous lines of therapies. In addition, dosing patterns and dose modifications for these therapies as well as clinical conditions while being treated were assessed.

Materials and Methods Study Design and Study Population

This was a retrospective, longitudinal cohort study using data from EMRs and a clinical database at Dana-Farber Cancer Institute. The center's Institutional Review Board granted approval for this study. Data for this study included longitudinal information on patient demographic and disease characteristics, treatment types, oncology-specific evaluations for tumor response, survival, and clinical conditions while on treatment as documented in physician notes.

Eligible patients had a biopsy-confirmed adenocarcinoma of the colon or rectum when at least 18 years of age, had confirmed metastatic disease, had initiated FTD/TPI or regorafenib treatment for mCRC, and had at least one visit at the study center. As depicted in supplemental online Figure 1, the index date was defined as the date of prescription of the first of FTD/TPI or regorafenib (index therapy). The baseline period was defined as the 6-month period preceding the index date. The follow-up period was defined as the time from the index date to the end of data availability (earliest of death, last visit at center, discharge to hospice, or end of study).

Data on demographic and clinical characteristics during the baseline period were collected and included primary cancer site, sites of metastases, genetic mutation status, comorbidities, and Likert pain score, which is based on a 10-point scale with 0 indicating no pain, 5 indicating moderate pain, and 10 indicating worst pain. Data also included treatment prior to the index therapy such as prior surgery, number of prior treatment regimens, and type of systemic therapies for mCRC. Treatment patterns in the follow-up period were assessed, including patient and clinical setting characteristics associated with index therapy, the line of the index therapy, dosing and dose modifications of the index therapy, and treatments after index therapy. Clinical endpoints as available in the EMR and clinical database were assessed in the follow-up period and included real-world best tumor response during the course of treatment with index therapy, which was based on information from both clinician assessment from notes as well as radiographic assessments from imaging reports. Best tumor response assessments during the treatment period were used to compute real-world overall response rate (rwORR) (i.e., proportion of patients who responded) and real-world disease control rate (rwDCR) (i.e., proportion of patients who had responded or a stable disease response at least 6 weeks after the index date). This approach differs from tumor response assessments in clinical trials that are performed at regular intervals. However, this real-world best tumor response assessment approach has been used in other observational settings [1, 7, 8]. OS was defined as the time between index date and death. The incidence of select clinical conditions while on treatment (e.g., hand-foot syndrome, rash or desquamation, general pain, fatigue, weight loss, increased aspartate aminotransferase [AST], increased alkaline phosphatase [ALP], hyperbilirubinemia, lymphocytopenia, neutropenia, thrombocytopenia, and febrile neutropenia) was assessed after the initiation of FTD/TPI or regorafenib therapy. Dose intensity was calculated as the mean daily dose of active treatment days. The relative dose intensity was defined as the ratio of dose intensity to recommended dose where recommended dose is 35 mg/m2 twice daily for FTD/TPI and 160 mg daily for regorafenib.

Statistical Analysis

The treatment groups of interest in the study were identified for analysis based on the index therapy (i.e., whether the patient first received FTD/TPI or regorafenib following an mCRC diagnosis). Descriptive statistics were calculated using frequencies and proportions for categorical variables and means, SDs, and medians for continuous variables. Differences between patients by index therapy were compared using Pearson χ2 tests (or Fisher's exact tests) for categorical variables, while continuous variables were compared using Wilcoxon rank-sum tests. Logistic regression models adjusting for potential baseline confounders that were prespecified a priori (i.e., gender, age, baseline Likert pain score, time from mCRC diagnosis to start of therapy, line of therapy) were performed to report the odds ratio (OR) and 95% confidence intervals (CIs) for rwORR and rwDCR.

Median OS and survival rates for each month after index therapy initiation were calculated using Kaplan-Meier analysis, in which patients who did not have an event were censored at the date of last follow-up; the log-rank test was employed to determine whether survival rates were different between patients who first initiated with FTD/TPI versus regorafenib as their index therapy. Cox proportional hazards models adjusting for potential baseline confounders that were prespecified a priori (i.e., gender, age, baseline Likert pain score, time from mCRC diagnosis to start of therapy, line of therapy) were used to compare OS between patients initiating different index therapy; HR and 95% CIs were estimated. The incidence of each clinical condition was estimated using multivariable Poisson regression models adjusted for relevant demographic covariates.

Planned subgroup analyses for patients who initiated the index therapy as second- or third-line mCRC treatment were performed. All analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC).

Results

Between April 2012 and December 2017, 126 patients were prescribed FTD/TPI, and 95 patients were prescribed regorafenib as their index therapies. Baseline patient demographic and clinical characteristics are presented in Table 1. A higher proportion of patients treated with FTD/TPI had Eastern Cooperative Oncology Group performance status 1 (46.8% vs. 32.6%; p value = .033), and the mean Likert pain score was higher in the FTD/TPI arm (0.9 vs. 0.3; p value = .007) with a lower proportion of FTD/TPI patients reporting no pain compared with regorafenib patients (79.8% vs. 92.6%; p value = .009). A higher proportion of patients treated with regorafenib had a left-sided primary colon tumor (57.4% vs. 44.0%; p value = .049). Characteristics of index therapy and patients at initiation of index therapy are described in Table 2. Patients who were first prescribed regorafenib versus FTD/TPI were more likely to be treated with this therapy during the early years of the study. The median follow-up for patients first treated with FTD/TPI and regorafenib was 7.1 and 6.3 months, respectively. Similar proportions of patients in both groups had surgery prior to index therapy (FTD/TPI: 65.1% vs. regorafenib: 75.8%; p value = .086) and surgery for resection of primary cancer site (FTD/TPI: 61.9% vs. regorafenib: 73.7%; p value = .065). Median relative dose intensity was significantly higher for patients initiating treatment with FTD/TPI versus regorafenib (FTD/TPI: 1.0 vs. regorafenib: 0.8; p value <.001). Twenty percent more patients with FTD/TPI as index therapy had no dose modifications compared with patients with regorafenib as index therapy (84.0% vs. 64.1%; p value < .001). Patients in both groups had a median of 3.0 pharmacologic regimens prior to index therapy; preindex and postindex therapies were comparable in both groups (supplemental online Table 1). Among 63 patients with FTD/TPI as index therapy that had a postindex therapy, 16 (25.4%) switched to regorafenib and 13 (20.6%) had both chemotherapy and biological therapy. Among 40 patients with regorafenib as index therapy that had a postindex therapy, 12 (30.0%) switched to FTD/TPI and 8 (20.0%) had a chemotherapy and biological therapy.

Table 1. Demographic and clinical characteristics among patients with FTD/TPI and regorafenib as index therapy Demographic characteristics FTD/TPI, n = 126 Regorafenib, n = 95 p value Age at index date, mean ± SD (median), yr 56.6 ± 11.1 (55.0) 58.7 ± 11.7 (57.0) .190 Race/ethnicity, n (%) White 107 (84.9) 85 (89.5) .321 Black 5 (4.0) 3 (3.2) .999 Hispanic or Latino 3 (2.4) 3 (3.2) .999 Asian/Pacific Islander 6 (4.8) 2 (2.1) .471 Unknown/not sure 5 (4.0) 2 (2.1) .702 Gender, n (%) Female 69 (54.8) 43 (45.3) .162 Male 57 (45.2) 52 (54.7) .162 Clinical and disease characteristics Body mass index, mean ± SD (median), kg/m2 27.2 ± 6.5 (26.2) 28.3 ± 7.2 (25.9) .235 ECOG performance status, n (%) 0 Fully active 34 (27.0) 24 (25.3) .773 1 Restricted in physically strenuous activity 59 (46.8) 31 (32.6) .033a 2 Require bed rest during <50% of waking day 5 (4.0) 3 (3.2) 1.000 3 Require bed rest during ≥50% of waking day 1 (0.8) 1 (1.1) 1.000 4 Completely disabled 0 (0.0) 0 (0.0) Unknown 27 (21.4) 36 (37.9) .007a Likert pain score available, n (%) 124 (98.4) 94 (98.9) Mean ± SD 0.9 ± 2.2 0.3 ± 1.3 .007a Primary cancer site, n (%) 125 (99.2) 94 (98.9) Colon 98 (78.4) 80 (85.1) .208 Left side 55 (44.0) 54 (57.4) .049a Right side 38 (30.4) 24 (25.5) .429 Side unspecified 5 (4.0) 2 (2.1) .701 Rectum 27 (1.6) 13 (13.8) .141 Rectosigmoid 0 (0.0) 1 (1.1) .429 Number of metastases at mCRC diagnosis, mean ± SD (median) 2.1 ± 1.1 (2.0) 2.0 ± 0.9 (2.0) .750 Genetic mutation status, n (%) KRAS 90 (71.4) 61 (64.2) .916 Mutated 48 (53.3) 32 (52.5) Wild type 42 (46.7) 29 (47.5) NRAS 22 (17.5) 2 (2.1) 1.000 Mutated 5 (22.7) 0 (0.0) Wild type 17 (77.3) 2 (100.0) BRAF 71 (56.3) 49 (51.6) .647 Mutated 4 (5.6) 1 (2.0) Wild type 67 (94.4) 48 (98.0) a p value <.05. Abbreviations: ECOG, Eastern Cooperative Oncology Group; FTD/TPI: trifluridine/tipiracil; mCRC, metastatic colorectal cancer. Table 2. Characteristics of index therapy treatment Treatment characteristics FTD/TPI, n = 126 Regorafenib, n = 95 p value Year of treatment initiation, n (%) 2012 1 (0.8) 16 (16.8) <.001a 2013 1 (0.8) 22 (23.2) <.001a 2014 0 (0.0) 27 (28.4) <.001a 2015 44 (34.9) 14 (14.7) <.001a 2016 49 (38.9) 7 (7.4) <.001a 2017 31 (24.6) 9 (9.5) .004a Time from mCRC diagnosis to start of therapy mean ± SD (median), mo 34.0 ± 25.1 (27.8) 31.3 ± 19.4 (29.2) .863 Follow-up time from index therapy mean ± SD (median), mo 9.2 ± 6.7 (7.1) 8.5 ± 7.6 (6.3) .157 Treatment setting, n (%) Clinical trial 4 (3.2) 0 (0.0) .136 Standard of care 116 (92.1) 95 (100.0) .006a Compassionate use 5 (4.0) 0 (0.0) .072 Unknown 1 (0.8) 0 (0.0) Line of index therapy, n (%) 2nd 8 (6.3) 4 (4.2) .487 3rd 39 (31.0) 29 (30.5) .946 4th 36 (28.6) 22 (23.2) .365 5th 26 (20.6) 22 (23.2) .652 6th plus 17 (13.5) 18 (18.9) .271 Patients with dosing information, n (%) 119 (94.4) 92 (96.8) First prescription dose, mean ± SD (median), mg/day 118.9 ± 27.3 (120.0) 142.2 ± 76.5 (160.0) Dosing patterns, patient level Dose intensity, mean ± SD (median), mg/day 117.5 ± 25.9 (120.0) 132.3 ± 66.9 (120.0) .010a Label recommended dose, mean ± SD (median), mg/day 130.8 ± 20.0 (130.4) 160.0 ± 0.0 (160.0) Relative dose intensity, mean ± SD (median) 0.9 ± 0.2 (1.0) 0.8 ± 0.4 (0.8) <.001a Patients with no dose modifications, n (%) 100 (84.0) 59 (64.1) <.001a Dose modifications per patient, mean ± SD 0.2 ± 0.5 0.4 ± 0.6 .001a a p value <.05. Abbreviations: FTD/TPI, trifluridine/tipiracil; mCRC, metastatic colorectal cancer.

Best real-world tumor response results are presented in Table 3. One hundred twenty (95.2%) patients with FTD/TPI as index therapy and 76 (80.0%) of those with regorafenib as index therapy had a clinician reported tumor response during the treatment period. Patients with FTD/TPI versus regorafenib as index therapy were more likely to respond to treatment (rwORR 52.5% vs. 34.2%; p value = .012) and achieve disease control (rwDCR 64.2% vs. 46.1%; p value = .013) based on best tumor response during the treatment period. Among patients who initiated index therapy as second- or third-line therapy, patients with FTD/TPI versus regorafenib were more likely to respond to treatment (rwORR 54.8% vs. 25.9%; p value = .018) and achieve disease control (rwDCR 69.0% vs. 37.0%; p value = .009). In the adjusted analyses, patients treated with FTD/TPI as index therapy had better tumor response and disease control compared with patients treated with regorafenib (rwORR OR, 2.57; 95% CI, 1.38–4.80; and rwDCR OR, 2.52; 95% CI, 1.36–4.68). Similarly, in the subgroup analyses of patients treated with index therapy as only second- or third-line mCRC treatments, patients treated with FTD/TPI had better tumor response and disease control compared with patients treated with regorafenib (rwORR OR, 4.11; 95% CI, 1.32–12.79; and rwDCR OR, 4.86; 95% CI, 1.57–15.04). Among all patients, most patients treated with FTD/TPI and regorafenib discontinued treatment because of disease progression (FTD/TPI: 85.2% vs. regorafenib: 77.9%; p value = .162). Patients treated with FTD/TPI versus regorafenib had less discontinuation because of toxicity/intolerance (8.2% vs. 24.2%; p value < .001).

Table 3. Unadjusted real-world tumor response among patients with FTD/TPI vs. regorafenib as index therapy Unadjusted real-world tumor response All patients Among patients in 2L and 3L of treatment FTD/TPI, n = 126 Regorafenib n = 95 p value FTD/TPI, n = 47 Regorafenib, n = 33 p value Patients with physician assessed tumor response, n (%) 120 (95.2) 76 (80.0) 42 (89.4) 27 (81.8) Real-world best tumor responses Real-world overall response rate, n (%) 63 (52.5) 26 (34.2) .012a 23 (54.8) 7 (25.9) .018a Stable disease 30 (25.0) 22 (28.9) .542 11 (26.2) 9 (33.3) .523 ≥6 weeks after index date 14 (11.7) 9 (11.8) .970 6 (14.3) 3 (11.1) .999 Progressive disease 27 (22.5) 28 (36.8) .029a 8 (19.0) 11 (40.7) .049a Real-world disease control rate, n (%) 77 (64.2) 35 (46.1) .013a 29 (69.0) 10 (37.0) .009a a p value <.05. Abbreviations: 2L, second line; 3L, third line; FTD/TPI, trifluridine/tipiracil.

Among patients who had FTD/TPI as their index therapy, 95 (75.4%) died (91 because of colon or rectal cancer) compared with 84 (88.4%) (79 died because of colon or rectal cancer) among those with regorafenib as their index therapy (p value = .015). Median OS was 7.5 months (95% CI, 6.0–8.8) versus 7.1 months (95% CI, 5.0–8.2) for all patients treated with FTD/TPI versus regorafenib. In the subgroup analyses of patients treated with index therapy as only second- or third-line mCRC treatments, patients treated with FTD/TPI versus regorafenib had a median OS of 7.7 months (95% CI, 4.4–11.5) versus 5.1 months (95% CI, 2.9–7.9), which was not significantly different. However, at each month of follow-up for the first 6 months, monthly survival was greater among patients who had initiated FTD/TPI versus regorafenib in the overall population and in the subgroup analyses of patients who initiated index therapy as only second- or third-line mCRC treatments. Specifically, patients who initiated FTD/TPI versus regorafenib had significantly greater survival rates at months 3 and 4 in the overall population and at months 2, 3, and 4 in the subgroup (Table 4). In the adjusted OS analysis controlling for gender, age, baseline Likert pain score, time from mCRC diagnosis, and index therapy after third-line of therapy in Table 5, patients treated with FTD/TPI versus regorafenib had similar risk of death (HR, 0.80; 95% CI, 0.59–1.09). In the subgroup analyses of only second- or third-line mCRC treatments, patients treated with FTD/TPI versus regorafenib had similar risk of death (HR, 0.60; 95% CI, 0.36–1.02).

Table 4. Kaplan-Meier estimates of survival among patients with FTD/TPI vs. regorafenib as index therapy Time period after index date Proportion of patients that survived All patients Among patients in 2L and 3L of treatment FTD/TPI, n = 126 Regorafenib, n = 95 p value FTD/TPI, n = 47 Regorafenib, n = 33 p value 1 month 98.4% 94.7% .126 97.9% 97.0% .808 2 months 92.0% 84.1% .071 95.7% 78.8% .021a 3 months 86.2% 73.4% .016a 84.6% 66.7% .045a 4 months 79.6% 65.8% .017a 75.6% 54.5% .034a 5 months 64.2% 60.1% .275 58.7% 51.5% .275 6 months 58.9% 55.5% .313 51.1% 48.5% .420 Median OS (95% CI), mo 7.5 (6.0–8.8) 7.1 (5.0–8.2) .3