Cost-Effectiveness Analysis of Genomic Proling in Early Breast Cancer in Colombia

Background: In Colombia, the best strategy to establish indication for adjuvant chemotherapy in early breast cancer (EBC) remains unknown. This study aimed to identify the cost-effectiveness of various strategies to establish the necessity of adjuvant chemotherapy. Methods: This study used an adapted decision-analytic model to compare cost and outcomes of care that includes Oncotype DX™ (ODX) or Mammaprint™ (MMP) tests with routine care without ODX or MMP tests (adjuvant chemotherapy for all patients) over a 5-year time horizon from the perspective of the Colombian National Health System (NHS; payer). Data were obtained from published literature and clinical trial database. The study population comprised women with hormone-receptor-positive (HR+), HER2-negative, lymph-node-negative (LN0) EBC with high-risk clinical criteria for recurrence. The outcome measures were incremental cost-effectiveness ratio (ICER; 2019 United States dollar per quality-adjusted life-year [QALY] gained) and net monetary benet (NMB). Results: ODX increases QALYs by 0.05 and MMP by 0.03 with savings of $2,445 and $570 compared with the standard strategy, respectively. ICER for ODX was −$41,857 and that for MMP was −$18,253 per QALY; NMB was $2,821 and $771, respectively. Both tests were cost-effective under the dened threshold. When the two tests were compared, ODX was more cost-effective than MMP. Sensitivity analysis revealed that with a threshold of 1 gross domestic product per capita, ODX will be cost-effective in 95.5% of the cases compared with 70.2% cases involving MMP. Probabilistic sensitivity analysis revealed that ODX was a consistently superior strategy. Conclusions: Genomic proling using ODX or MMP tests to dene the need of adjuvant chemotherapy treatment in patients with HR+ and HER2− EBC is a cost-effective strategy that allows Colombian NHS to maintain budget.


Background
Breast cancer tumors are the most frequently occurring type of tumors; this cancer type is the leading cause of cancer-related death in women in Colombia, with annual incidence of 48.3 and mortality of 13.1 per 100,000 women (1). Adjuvant chemotherapy reduces the risk of recurrence and increases overall survival; however, it involves a risk of associated toxicity, decreased quality of life, and a signi cant burden on the healthcare system (2)(3)(4). The risk of recurrence determines a patient's eligibility for adjuvant chemotherapy. Several clinicopathological parameters are considered to determine the risk of recurrence, such as lymph node involvement, tumor size, subtype and histological grade, lymphovascular invasion, proliferation markers, hormone-receptor (HR) status, and HER2/neu (5,6). Adjuvant chemotherapy may not be recommended in patients with HR-positive (+) and HER2-negative (−) early breast cancer (EBC) (2). In this population, other clinical criteria, such as tumor size, the degree of differentiation, and patient age, can be taken into consideration (7). However, recommendation based solely on clinical criteria may lead to a signi cant proportion of these patients being exposed to the adverse effects of chemotherapy and deterioration of quality of life with limited or uncertain bene t (8).
Recently, quantitative tests evaluating gene expression using microarray and reverse-transcription polymerase chain reaction techniques were developed to better establish the prognosis of patients with EBC with an HR + and HER2 − status. Two of these techniques are Oncotype DX™ (ODX; Genomic Health, Redwood City, CA) and Mammaprint™ (MMP; Agendia, Irvine, CA) (9,10). Both techniques have been validated in different studies and proven more accurate at estimating the recurrence risk compared with clinical parameters and other algorithms such as Adjuvant! Online and Nottingham Prognostic Index (NPI) (11)(12)(13). Furthermore, there are no relevant variations in test results among genetically different populations such as Latin-American patients (14). In Colombia, according to guidelines, recommendations, and clinical practice, the decision about adjuvant chemotherapy in patients with HR + and Her2 − EBC is based on clinical and pathologic characteristics, and Oncotype DX™ (ODX) or Mammaprint™ (MMP) are recommended and approved for the support this decision. (15). The costs of performing ODX or MMP are assumed by Colombian National Health System (NHS).
Direct costs of care for patients with breast cancer are high, and a signi cant proportion of these are incurred from chemotherapy, which includes treatment of its side effects. Therefore, a strategy that allows an adequate estimate of the risk of recurrence and identi cation of patients who will bene t from adjuvant chemotherapy will facilitate healthcare cost reductions. This study aimed to determine which of the following interventions was more cost-effective for patients with HR + and HER2 − EBC with high-risk clinical criteria de ned as that used in the Microarray in Node-Negative Disease May Avoid Chemotherapy trial:(16) administer adjuvant chemotherapy in all patients (standard intervention) or de ne treatment indication based on results from ODX or MMP genomic tests.

Overview
An economic cost-effectiveness study evaluating the performance of ODX or MMP test was performed in a cohort of patients with HR + and HER-2 − EBC, without lymph node involvement (LN0) and with high risk of recurrence de ned by clinical criteria. Table 1 summarizes the scope of our model. Following the suggestions of the World Health Organization, we compared the incremental cost-effectiveness ratio (ICER) with 1 gross domestic product (GDP) per capita and 3 GDP per capita, until more precise threshold estimates become available in Colombia (17,18). The GDP per capita data were obtained from the Banco de la República reports for the year in which the evaluation was performed (2019) (19).

Resource Use and Cost
The costs were reported as the value of 2019 USD (1 USD = 3,281 Colombian pesos, according to an average of 2019 COP market exchange rate obtained from Banco de la República reports) (27), and only direct costs were accounted for.
The following sources were used to estimate costs: (1) In MMP and ODX genomic tests, the prices in the Colombian market (for 2019) consulted with the providers were used and (2) cost-generating events and frequency of use for each state were identi ed through a) clinical recommendations from clinical practice guidelines and b) a panel of experts. Supplementary Table 1 in Additional le 1 summarizes the resources, cost, and references used.
In this study, only febrile neutropenia as adverse event of chemotherapy was taken into account considering that it is the most frequent severe adverse event of adjuvant chemotherapy in breast cancer (28-30).
A type or base case was designed. This type case involved the use of cost-generating events (i.e., consultations, medications, diagnostic tests, hospitalization, and procedures). Subsequently, the monetary cost for the Colombian NHS of each of the cost-generating events was included.

Model Description
The model structure was developed by consensus among the authors, and it took into account previous published and validated models (31). The model followed the clinical pathway that is in accordance with the recommendations of the clinical practice guidelines under the assumption that it corresponds to the best available description of usual clinical practice in Colombia (15). An analytical decision model was developed, in which a cohort of patients with HR+, HER2−, and LN0 EBC with high-risk clinical criteria was assigned to the standard intervention in which all patients received chemotherapy or underwent ODX or MMP and, according to the results, were classi ed as high or low risk.
For patients assigned to the standard intervention (chemotherapy for all), distant recurrence-free survival (DRFS) and overall survival (OS) were estimated according to the information available from the TAILORx trial database (NCT00310180).
From this database, we selected the population that met the high-risk clinical criteria, i.e., a tumor size (T) of > 3 cm or > 2 cm and a degree of differentiation of 2 or a T of > 1 cm and a degree of differentiation of 3, according with the criteria used in MINDACT trial (16). DRFS and OS were estimated for patients who met these criteria, and they subsequently underwent chemotherapy.
In ODX, this same database was considered, and patients were categorized as high or low risk according to the de nitions of the TAILORx trial (20), where high risk corresponds to a score of ≥ 16 for patients aged ≤ 50 years or a score of ≥ 26 for those aged > 50 years and low risk corresponds to a score of ≤ 15 for patients aged ≤ 50 years and ≤ 25 for those aged > 50 years. DRFS and OS were similarly estimated. In MMP, the possibility of a high-or low-risk classi cation was considered, and the risk of recurrence and death for each risk category was based on the report of the MINDACT trial in a population with negative nodes (N0 Similarly, the effect on QALY in patients undergoing chemotherapy was based on literature reports (25).

Model Assumptions
The most important assumption in our model is the estimation made for QALYs and utilities; however, the process was done according to the recommendations for the country in this regard (17). By contrast, for the population receiving standard intervention (chemotherapy), the estimates were made from the database of a clinical study that might not re ect the conditions of routine clinical practice and there may be differences according to patient races. However, both clinical characteristics and treatments that these patients received were similar to those recommended in different clinical practice guidelines and were contrasted with the panel of experts (15). Furthermore, the database used for modeling contained Latin population data. It was assumed that there was 100% adherence to the result of ODX or MMP-if the result corresponded to a high-risk status, the patient was assumed to require chemotherapy and, if low risk, no chemotherapy was needed. However, this trend has been reported in a lower percentage in a previous case series (32).

Statistical Analysis
Deterministic sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA) were performed. A univariate DSA was developed for the assumptions, probabilities, and costs, whereas other variables of the model were xed. A tornado diagram was used to show the effect of the variation of these parameters on the incremental NMB. To evaluate the uncertainty of any parameter, a PSA was performed using the Monte Carlo simulation method. The simulation was applied 1,000 times to ensure the reproducibility of the model. Table 2 enlists the variables used as inputs for our model as well as their values, ranges, and distribution parameters applied in DSA and PSA. For the population who underwent chemotherapy (standard intervention), the estimated 5-year DRFS rate was 94% and the OS rate was 97%. Supplementary Fig. 1 Fig. 2 in Additional le 1).
In addition, the two genotyping strategies represented an incremental NMB of $2,821 for ODX and $771 for MMP, and this was considered when a willingness to pay (WTP) threshold of 1 GDP per capita was considered the base. When two ODX tests were compared, ODX was more cost-effective than MMP. Table 3 summarizes the aforementioned ndings.

Sensitivity Analysis
Regarding DSA, variables that were considered at the discretion of the researchers and the panel of experts were de ned as having greater uncertainty for the model, which included a) utilities during adjuvant chemotherapy, recurrence, and drug-free period recurrence and b) costs for chemotherapy, adverse events, treatment to recurrence, and palliative care. Supplementary Fig. 3 in Additional le 1 shows the tornado diagram for the DSA for ODX in QALY.
In our analysis, the variable with the greatest in uence was the monthly cost of adjuvant chemotherapy.
Notably, we observed that the variation in utility values did not have a signi cant effect. This suggested that despite the limitations imposed by their estimation from populations other than that of Colombia, they do not signi cantly affect our ndings.
Similar ndings were obtained in the DSA for MMP ( Supplementary Fig. 4 in Additional le 1). Given that chemotherapy treatment costs <$274, MMP may not be cost-effective. In the PSA, considering that a threshold of 1 GDP per capita for 2019 corresponded to $6,428, ODX had 95.5% probability of being costeffective and MMP had 70.2% compared with the standard strategy (chemotherapy for all patients).
Comparison of the two tests revealed that ODX had 99.1% probability of being more cost-effective than MMP. Figure 3

Discussion
The use of adjuvant chemotherapy in EBC has been bene cial to some patients by decreasing tumor recurrence and increasing OS; however, in certain cases, chemotherapy can generate serious adverse events that cause deterioration in the quality of life and substantial increase in health costs (2,4,24).
Furthermore, in this population, it is typically di cult to determine whether chemotherapy should be administered (35).
Gene expression pro les such as MMP and ODX have established more precisely the prognosis and helped de ne the bene t of chemotherapy treatment in an individual assessment. This results in the accurate selection patients and avoids unnecessary therapies; however, performing these tests may imply an additional cost (16,20). These tests are recommended by different clinical practice guidelines, especially for patients with a high clinical risk, and they are often used in Colombian oncology practice (15).
In this study, which was possibly the rst in Colombia, we found that performing ODX or MMP is a costeffective strategy for the health system and generates economic savings. In our model, although we observed an increase in life-years after comparing the performance of standard strategy (4.36) with that of MMP (4.34) or ODX (4.33), these results are at the expense of deterioration in the quality of life that can be attributed to chemotherapy treatment, resulting in a bene t in terms of QALY that favors MMP or ODX (0.03 and 0.05, respectively). Despite the possibility of these differences in improvements in both life-year and QALY not being statistically or clinically signi cant, the capability of MMP or ODX tests to better select patients for chemotherapy results in the reduction of the cost of chemotherapy treatment. The incremental NMB for ODX is $2,741 and that for MMP is $771, indicating that genomic pro ling using these tests generates an economic surplus compared with the standard strategy in the WTP threshold de ned for Colombia as 1 GDP per capita for the QALY analysis. In this case, the cost to obtain the bene t is less than the maximum amount that the Colombian NHS would consider paying for this bene t. In the sensitivity analysis, MMP could not be considered economically acceptable only when the costs of adjuvant chemotherapy were <$274 per month of treatment.
Notably, even when the system is not willing to pay any cost for this bene t, i.e., with a WTP threshold of $0, both tests are cost-effective (incremental NMB for ODX at $2,446 and incremental NMB for MMP at $570).
From the perspective of the Colombian NHS, for a WTP of 1 GDP per capita, there is a 95.5% and 70.2% probability that ODX and MMP tests will be cost-effective, respectively. Importantly, even with a WPT of COP 0, the probability that the tests are cost-effective is high, especially for one of the tests (99.1% for ODX and 66.7% for MMP).
Our results are in agreement with those reported previously in the literature. A cost-effectiveness study of the genomic pro le for breast cancer conducted in Canada that included information from the TAILORx (20) and MINDACT (16) trials showed that the genomic pro ling of breast cancer patients using ODX or MMP tests is a cost-effective strategy below the threshold of WTP de ned for this study when compared with the standard management, i.e., the absence of any test. In this analysis, ODX has a 89.2% probability and MMP has 89.2% probability of being cost-effective for a WTP threshold of Canadian dollar 50,000 (36).
Several systematic reviews have concluded that performing the genomic pro le in EBC to de ne adjuvant chemotherapy treatment is a cost-effective strategy (37)(38)(39). However, other analyses have shown that this nding is not consistent in all population subgroups and that ODX genomic pro le is cost-effective when performed in a high-clinical risk population and not in a low-clinical risk population. A costeffectiveness study conducted by the UK National Institute for Excellence in Health and Care found that neither ODX nor MMP was cost-effective from the perspective of the UK health system (40). No predictive role of the bene t of chemotherapy was considered for ODX or MMP. Notably, at the time of performing this analysis, the results of the TAILORx trial were unknown, which demonstrated the ability of ODX to establish not only prognosis but also the bene t of adjuvant chemotherapy (20). When the predictive role of ODX to establish the bene t of adjuvant chemotherapy was included, this test was cost-effective, particularly for patients with high clinical risk (for this study de ned as the subgroup with NPI > 3.4) a nding similar of ours (40). In MMP, despite the results of the MINDACT trial (16), this test was not costeffective (40).
Several cost-effectiveness studies have not considered this analysis for clinical risk subgroups, which could favor the new test as cost-effective, and for this reason, the incorporation of clinical characteristics into the cost-effectiveness models is recommended (41). In contrast, the performance of MMP is only recommended in patients with high clinical risk, and the combination of these clinical criteria with ODX result can increase its prognostic capacity (42,43). Only patients with high-risk clinical criteria were included in our model, which is a conservative strategy, demonstrating that performing MMP or ODX in this population is a cost-effective strategy. Hall et al. found results similar to ours in the United Kingdom when they used a model that also included patients with high clinical risk with lymph node involvement (31). In addition to MMP and ODX, other tests such as PAM-50 (Prosigna™), MammaTyper™, IHC4, and IHC4-AQUA™ (NexCourseBreast™) were also evaluated by Hall et al. and found an 86% probability that gene expression pro les are cost-effective in de ning the need for adjuvant chemotherapy in patients with EBC (31).
In our study, when two ODX tests were compared, and they were more cost-effective than MMP with an incremental NMB of $2,050 and 99% probability of being more cost-effective. Our ndings suggest that to achieve results similar to those of ODX test, the costs of MMP test should be lower.
Most cost-effectiveness studies have assumed that the relative risk reduction (RRR) for distant recurrence attributed to chemotherapy varies according to different genomic risk groups, i.e., the RRR is 0 for patients with low genomic risk and higher for those with high genomic risk. These assumptions make genomic testing more cost-effective because it can better establish the magnitude of chemotherapy bene t than the traditional clinical criteria. However, the predictive values of these tests, for these cost-effectiveness studies, were based on limited information based on retrospective analysis (44,45 Considering that genomic pro ling using ODX and MMP tests is a cost-effective and cost-saving strategy, establishing the budget effect of these tests could be important to de ne whether they can be included in Colombia's health bene t plan.

Conclusions
Genomic pro ling using ODX and MMP tests is a cost-effective strategy for the Colombian NHS. This strategy generates savings in the health system when compared with the standard treatment strategy.
When these two genomic pro ling strategies are compared, ODX is a more cost-effective strategy than MMP and allows for greater savings. These tests should be indicated in a population with HR+, HER2 − EBC with high-risk clinical criteria.  Competing interests: The authors declare that they have no competing interests.

List Of Abbreviations
Funding: Not applicable.
Authors' contributions: LR conceived the project, wrote, and prepared the manuscript; MR-R: conceived the project, discussed, and reviewed the manuscript. DR and AFC: discussed and reviewed the manuscript. All authors read and approved the nal manuscript.