Effectiveness and Cost-Utility Analysis of Different Doses of Irinotecan Plus Bevacizumab in Patients With Metastatic Colorectal Cancer: A Long-Term and Prospective Cohort Study

Objective Patients with metastatic colorectal cancer (mCRC) had oncological benefits with irinotecan dose escalation of FOLFIRI regimen combined with bevacizumab according to UGT1A1 genotypes in our previous study. In the current study, we performed a quality of life (QOL) outcome evaluation and cost-utility analysis of different irinotecan dose regimens in patients with mCRC. Materials and Methods With inverse probability-of-treatment weighting (IPTW) matching on all covariates, 75 patients with dose escalation of irinotecan (study group) and 121 patients with the recommended dose of irinotecan (control group) were recruited between October 2015 and December 2019. The QOL outcome measures were Functional Assessment of Cancer Therapy-Colorectal, Beck Anxiety Inventory, Beck Depression Inventory, and SF-36; cost-utility outcome measures were medical direct costs, quality-adjusted life-years (QALYs), and incremental cost-utility ratios (ICURs). Results All mCRC patients exhibited a significant decrease in both emotional wellbeing and depression from pretherapeutic period to posttherapeutic 6th month (P < 0.05); however, from the posttherapeutic 1st year to the 2nd year, improvement in most QOL measures was significantly better in the study group than in the control group (P < 0.05). Over a 2-year time period, the study group had higher total medical direct costs than the control group (US$ 54,742 ± 14,013 vs. US$ 54,608 ± 9,673) and higher average QALYs gained (1.88 vs. 1.65), with an ICUR of US$ 583 per QALY gained. Conclusion For patients with mCRC, irinotecan dose escalation appeared cost-effective with considerable QOL improvements during the study period. Further randomized, multi-institutional controlled trials are warranted to corroborate these results.


INTRODUCTION
With the recent advances in pharmacogenomic era, patients undergo genetic testing to determine their genotype before treatment, based on which a suitable medication or dosage is administered (1)(2)(3). For metastatic colorectal cancer (mCRC) patients, the standard chemotherapy regimens are 5-fluorouracil (5-FU), leucovorin, and oxaliplatin (FOLFOX); oxaliplatin with capecitabine (CAPOX); or 5-FU, leucovorin, and irinotecan (FOLFIRI). Bevacizumab is a recombinant humanized monoclonal antibody targeting the vascular endothelial growth factor and is usually added to these standard regimens to enhance the efficacy. Our previous multicenter, randomized, controlled, open-label trial study revealed that exceeding the recommended irinotecan dose is safe and effective when a regimen of FOLFIRI plus bevacizumab is administered in mCRC patients with UGT1A1 * 1/ * 1 and UGT1A1 * 1/ * 28 genotypes. Additionally, pretherapeutic UGT1A1 genotyping-guided dose adjustment achieved better outcomes compared to the standard regimen. This intention-totreat (ITT) analysis also demonstrated the benefits of escalating doses of irinotecan under UGT1A1 genotyping. Likewise, other previous studies have also suggested that higher doses of irinotecan in the first-line or later-line setting in mCRC can confer favorable clinical outcomes (4-7). Roncato et al. estimated the per-patient cost of treating irinotecan-related toxicity associated with the UGT1A1*28 patient genotype. Mean per-patient cost was €812 for UGT1A1*1/*1 patients, €1,119 for UGT1A1*1/*28 patients, and €4,886 for UGT1A1*28/*28 patients (8). Notably, the incremental treatment cost per patient was €4,074 higher in UGT1A1*28/*28 patients compared to UGT1A1*1/*1 patients and €307 higher in UGT1A1*28/*28 patients compared to UGT1A1*1/*28 patients. Kristin et al. also found that the total per-patient cost of a FOLFOX or FOLFIRI chemotherapy regimen was 359 million Indonesian Rupiah (IDR) with a mean average of 1.90 quality-adjusted life-years (QALYs) over a life-time horizon. Additionally, although FOLFOX or FOLFIRI plus bevacizumab increased the total per-patient cost by 108 million IDR, it increased QALYs by 0.17 (9).
Although irinotecan dose escalation demonstrates a more optimal treatment effect in patients with mCRC undergoing UGT1A1 genotyping and receiving FOLFIRI, the cost utility of different doses of irinotecan plus bevacizumab has not been comprehensively investigated. Additionally, most studies on the economic and cost-effectiveness of this treatment have analyzed the short-term results; however, few studies have analyzed the longterm results. Therefore, the objective of this prospective, long-term follow-up study was to perform a quality of life (QOL) outcome evaluation and a cost-utility analysis of different doses of irinotecan plus bevacizumab in patients with mCRC in the first-line setting.

Study Design and Study Population
The study population included patients with mCRC at a medical center between October 2015 and December 2019. The inclusion criteria were as follows: (a) patients who underwent UGT1A1 genotyping before treatment; (b) mCRC patients with histologically diagnosed adenocarcinoma, and (c) patients aged 20-80 years. The exclusion criteria were as follows: (a) no receipt of bevacizumab plus FOLFIRI, (b) the UGT1A1 genotype 7TA/ 7TA (UGT1A1*28*28); (c) <20 or >80 years of age; (d) no dose esclation due to discomfort experienced by those in the dose escalation group; (e) being pregnant or breastfeeding; (f) a major comorbidity; and (g) having not signed the consent form. The recommended irinotecan dose in the FOLFIRI regimen is 180 mg/m 2 based on a dose-finding study (2,4,10). The detailed treatment regimen on irinotecan dose escalation including patient withdrawal was described in our previous study protocol (11). According to the above criteria, mCRC patients were divided into two groups: 75 patients were categorized into the dose escalation group (>180 mg/m 2 , study group) and 121 patients into the recommended dose group (180 mg/m 2 , control group). The patients in the recommended dose group received irinotecan at a dose of 180 mg/m 2 . Figure 1 shows a flow diagram of the current study illustrating the enrollment, allocation, and matching analysis processes. Before study initiation, approval was obtained from the Institutional Review Board of our hospital (KMUHIRB-(EI)-20150147). The patients/ participants provided their written informed consent to participate in this study.

Quality of Life Measures
The QOL measures included the Functional Assessment of Cancer Therapy-Colorectal (FACT-C) (12), Beck Depression Inventory (BDI) (13), Beck Anxiety Inventory (BAI) (14), and SF-36 (15). Higher FACT-C and SF-36 scores indicated better outcomes, but higher BDI and BAI scores exhibited worse outcomes. All enrolled patients were scheduled to complete all the above assessments at four time points: pretherapeutic and therapy at posttherapeutic 6th month, 1st year, and 2nd year.

Cost Estimation
In accordance with the reimbursement criteria established by the National Health Insurance Administration, medical cost structure included total medical direct costs during therapy and total inpatient costs, total outpatient costs, and total medical emergency costs at posttherapeutic 2 years after discharge. All medical direct costs included expenses for physicians, ward, pharmacy, laboratory, inspection, medical materials, etc. All cost inputs were adjusted to US$ 2019 according to the consumer price index (CPI) and discounted annually by 3%.

Estimation of Utility
To estimate QALYs, cost-utility analysis often uses "utility scores" (health state valuations) anchored by 0 and 1, where 0 indicates death and 1 indicates full health. This study used the time trade-off valuation procedure to convert EuroQol-5D (EQ-5D) total scores to utility scores (16). The utilities reported by each patient were multiplied by the assumed duration of sustained benefit after intervention (summed up to the end of 2 years) to estimate QALYs. To maintain consistency with QALYs calculation, this study assumed that the only resources used by patients were those captured during the 2 years of followup. That is, the analysis assumed that patients did not incur any other healthcare costs during the remainder of the year.

Statistical Analysis
In the main patient-level analysis, we used inverse probability-oftreatment weighting (IPTW) based on propensity score to construct a weighted cohort of patients with different doses of irinotecan plus bevacizumab but similar with respect to other study characteristics. IPTW based on propensity scores was used to account for differences in study characteristics between the two regimens (17). The resulting differences between the weighted groups of interest reflected the average treatment effect ( Table 1). The distribution of FACT-C scores at each time point was analyzed in terms of median, range and interquartile range and visualized by box-and-whisker plots.
A common limitation of longitudinal studies is the absence of an appropriate statistical methodology that can control for censoring and intercorrelations that occur when measures are repeatedly obtained for the same pool of subjects. To address this limitation, this study used a generalized estimating equation (GEE) model to cluster mCRC patients; the GEE model was also useful for generating propensity scores. Moreover, total scores for each QOL measure were compared between the study group and control group by using the differences-in-differences (DID) methodology (i.e., a pre-post study design with a comparison group) (18). Standard errors in DID in the predicted values were estimated using the bootstrap technique (1,000 replications and sample sizes equal to the original sample size) (19). Additionally, effect size was obtained using Cohen's d statistics (i.e., the difference between the mean posttherapeutic QOL value and the mean pretherapeutic QOL value divided by the pooled standard deviation) (20).
After converting EQ-5D scores into utility scores, the number of QALYs over a period of 2 years was calculated for each patient by using the area under the curve approach after controlling for imbalances in baseline utility scores (21). The incremental costutility ratio (ICUR) was calculated as the ratio of the difference in mean cost per patient to the difference in mean QALYs per patient between the study and control groups. A willingness-to-pay (WTP) threshold of gross domestic product (GDP) US$ 26,263.5 per QALY was used to assess cost-effectiveness. A regimen is termed dominant when it is both less costly and more effective. To derive costeffectiveness acceptability curve, this study performed nonparametric bootstrapping on the incremental cost and effectiveness with 1,000 replications and a cost-effectiveness acceptability frontier. All P values reported were two-sided, and P values of <0.05 were considered statistically significant. The software package used to perform GEE in all statistical analyses was xtgee in Stata version 13.0 (Stata Corp LP, College Station, TX, USA), and a decision tree model was built using the TreeAge Pro 2017 (Tree-Age Software Inc. Williamstown, MA, USA).

RESULTS
To unify the study characteristics of the two groups and reinforce the subsequent research results, IPTW was used to match all study characteristics; thus, no variables were significantly different between the groups ( Table 1). Differences in the QOL Between the Two Groups After IPTW Table 4 presents a comparison of the differences in QOL measures between the two groups at four time points: pretherapeutic (T0), posttherapeutic 6th month (T1), posttherapeutic 1st year (T2), and posttherapeutic 2nd year (T3). Compared with the control group, at T0, the study group had a significantly better function in FACT-C social/family wellbeing (P = 0.024) and BAI (P = 0.039); at T1, the study group had a significantly better function in FACT-C functional wellbeing (P = 0.025), BAI (P = 0.003), SF-36 physical functioning (P = 0.008), vitality (P = 0.031), and role limitation due to emotional problems (P = 0.033); at T2, the study group had a significantly better function in FACT-C physical wellbeing (P = 0.025), FACT-C emotional wellbeing (P = 0.002), and SF-36 role limitation due to physical problems (P = 0.033); at T3, the study group had a significantly better function in FACT-C colorectal cancer-specific wellbeing (P = 0.017), BAI (P = 0.009), SF-36 physical functioning (P = 0.002), role limitation due to physical problems (P = 0.004), and role limitation due to emotional problems (P = 0.017). Overall, from posttherapeutic 1st year to the 2nd year, improvements in most QOL   measures were significant in the study group compared with those in the control group (P < 0.05).

Cost-Utility Analysis of the Study Group Compared With the Control Group
The mean total medical direct costs per patient over a 2-year time period included the mean total medical direct costs during therapy and the mean total medical direct costs at posttherapeutic 2 years after discharge. The mean total medical direct cost of the study group was US$ 54,742 (standard deviation, SD, US$ 14,013) compared with US$ 54,608 (SD US$ 9,673) for the control group, resulting in mean incremental costs of US$ 134 ( Table 5). Treatment in the study group led to a decrease in the mean utility score from 0.97 (SD 0.09) at the pretherapeutic time point to 0.95 (SD 0.09) at posttherapeutic 2nd year compared with the control group, which had a mean of 0.94 (SD 0.16) at the pretherapeutic time point to 0.86 (SD 0.24) at  posttherapeutic 2nd year ( Table 6 and Figure 3). After adjusting for pretherapeutic utility, the mean QALYs per patient in the study group was 1.88, whereas the mean QALYs per patient in the control group was 1.65. This resulted in an ICUR of US$ 583 per QALY over the first 2 years for patients in the study group.

Sensitivity Analysis
Probabilistic Sensitivity Analysis (PSA) The ICURs for the 1,000 samples in the PSA are shown in the scatter plot ( Figure 4). All points were under the US$ 26,263.5per-QALY level, and 100% of the tested ICURs were in the  northeastern quadrant. The cost-effectiveness acceptability curve is also shown in Figure 5 for varying values of WTP per QALY.

Univariable Sensitivity Analysis (USA)
The results of the USA are shown in the tornado diagram ( Figure 6). The parameters with the greatest influence on the ICUR were total costs in the study group with improved QOL, followed by total costs in the study group with maintained QOL, and total costs in the control group with maintained QOL. Even with a broad variation in range for each parameter, the ICUR remained below US$ 26,263.5 per QALY.

DISCUSSION
In both groups of patients with mCRC, physiological functions and emotional wellbeing significantly worsened at posttherapeutic 6th month compared with the pretherapeutic status. This can probably be attributed to the adverse effect of the chemotherapy. The study group exhibited significant improvement in physiological functions and emotional wellbeing from posttherapeutic 6th month to posttherapeutic 1st year. However, the control group exhibited worse emotional wellbeing, possibly because their physiological functions had not fully recovered. From posttherapeutic 1st year to posttherapeutic 2nd year, the study group gradually improved in most QOL measures. By contrast, the control group exhibited slow improvement in physiological functions, and their moods were still affected. These results were consistent with previous reports that compared patient conditions before chemotherapy and radiotherapy; the physiological functions, social functions, and nausea and vomit symptoms of cancer patients worsened during the 6-month treatment, leading to negative emotional functions (22)(23)(24). However, most QOL outcomes improved but they still were inferior to the initial status at 1-2 years after chemotherapy and radiotherapy. These results emphasized that baseline characterization of symptom status should be incorporated into clinical trials as they may affect symptom burden. Compared with the control group, the study group had higher mean total medical direct costs, possibly because of the dose difference. The study group had higher hospitalization expenses during therapy and subsequent hospitalization expenses      than those with UGT1A1*1/*28 and UGT1A1*1/*1 (10,25). Gradual irinotecan dose escalation according to UGT1A1 genotypes can yield better treatment outcomes and more easily prevent adverse effects. The present study indicated that escalated irinotecan dose can help achieve the optimal values for cost-effective incremental costs of care in Taiwan. Few studies have analyzed treatment with different irinotecan doses combined with bevacizumab. Obradovic et al. divided patients with mCRC into those with or without specific UGT1A1 genotypes (26); however, they did not compare patients who did not undergo genotyping (administered the fixed-dose regimen) and patients who underwent UGT1A1 genotyping (administered escalated-dose regimen). If patients with UGT1A1*1/*28 and UGT1A1*1/*1 were administered an escalated dose of irinotecan and those with UGT1A1*28/*28 were given a reduced irinotecan dose, the overall total medical direct costs may be reduced. Butzke et al. performed a sensitivity analysis and discovered that a UGT1A1 test before irinotecan chemotherapy and medical expenses were critical factors determining costs (27). Gold et al. conducted a sensitivity analysis and revealed that only when irinotecan dose adjustment achieved 98.4% of the effectiveness of the full dose, the DNA test can be continued (28). When ICUR was US$ 100,073 per QALY and WTP was US$ 100,000, UGT1A1 genotyping test remained the most optimal choice before administering irinotecan. Therefore, treatment in the study group resulted in fewer disease recurrences, and thus, an ICUR decline when additional costs after disease recurrences increased. Given this, the escalated dose of irinotecan will be cost-effective compared with the recommended dose of irinotecan when combined with other costlier treatment strategies in mCRC patients. To enable decision making for tailored therapy, before administering irinotecan, different doses should be determined according to the UGT1A1 genotype, thereby achieving more favorable cost-utility.
The following limitations of this study must be acknowledged. First, this was not a randomized study. At baseline, the study group (escalated dose group) had higher mean scores for FACT-C social/ family wellbeing and BAI than the control group (recommended dose group). Therefore, the magnitude of QOL improvements obtained by different doses of irinotecan plus bevacizumab may have been underestimated. Nonetheless, the potential bias would not have changed our result that the study group seemed to achieve more cost-effectiveness than the control dose group. Second, we did not include costs involved in the nonhealthcare sector or from the societal perspective, including productivity loss and out-of-pocket expenses. Third, repeated measures of QOL outcomes were limited to 2 years. Basic symptoms of pain, neuropathic pain, and chemicals were not assessed, which is a noted limitation of the quality of life assessment in this study. Further randomized clinical trials are needed to corroborate the benefit on clinical outcomes, additional costs other than healthcare costs including survival, and overall societal impact.
The improved effectiveness in the study group was corroborated by evidence of significant improvements in QOL outcomes from posttherapeutic 6th month to posttherapeutic 2nd year. Nevertheless, we suggest that before the introduction of irinotecan treatment, patients with mCRC undergo UGT1A1 genotyping to help determine the appropriate irinotecan dose, thereby achieving higher cost-utility and better medical resource allocation. Further scientific research is needed to determine what doses achieve the most desirable therapeutic effects in different UGT1A1 genotypes, which would translate into improved QOL after such a burdensome treatment of the underlying disease. Our results may serve as potential references for health departments, academic units, and medical supply units to treat mCRC.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the institutional ethics committee of our hospital (KMUHIRB-(EI)-20150147). The patients/participants provided their written informed consent to participate in this study.