Cost-Effectiveness of Adjuvant Treatment for Ductal Carcinoma In Situ

PURPOSE

Ductal carcinoma in situ (DCIS) accounts for 20% of breast cancer cases in the United States and is potentially overtreated, leading to high expenditures and low-value care. We conducted a cost-effectiveness analysis evaluating all adjuvant treatment strategies for DCIS.

METHODS

A Markov model was created with six competing treatment strategies: observation, tamoxifen (TAM) alone, aromatase inhibitor (AI) alone, radiation treatment (RT) alone, RT + TAM, and RT + AI. Baseline recurrence rates were modeled using the NSABP B17 and RTOG 9804 trials for standard-risk and good-risk DCIS, respectively. Relative risk reductions and adverse event rates for each treatment strategy were derived from meta-analyses of large randomized trials. We used a willingness-to-pay threshold of $100,000 in US dollars/quality-adjusted life-year and a lifetime horizon for two cohorts of women, age 40 and 60 years. Comprehensive sensitivity analyses evaluated the robustness of base-case results.

RESULTS

RT alone was cost-effective for patients with standard-risk DCIS, and observation was cost-effective for patients with good-risk DCIS, across both age groups. Strategies including TAM or AI resulted in fewer quality-adjusted life-years than observation, because of the prolonged decrement in quality of life outweighing the modest benefit in ipsilateral risk reduction. In sensitivity analysis, RT alone was cost-effective for age 40, good-risk patients when ipsilateral risk reduction matched that of the RTOG 9804 trial, there was minimal increased risk of contralateral breast secondary malignancy, or there was strong patient willingness to pursue RT.

CONCLUSION

Our findings suggest that cost-effective and clinically optimal treatment strategies are RT alone for standard-risk DCIS and observation for good-risk DCIS, with personalization on the basis of patient age and preference for RT. Hormonal therapy is likely suboptimal for most patients with DCIS.

INTRODUCTION

Breast cancer is both the most common cancer and the most expensive cancer to treat in the United States, with annual expenditures estimated at more than $20 billion in US dollars (USD).^1,2 Ductal carcinoma in situ (DCIS) accounts for 20% of all breast cancer cases, with close to 50,000 new cases annually—if classified independently, it would be the fifth most common malignancy in women.³ Not only does the treatment of DCIS consume a sizable portion of health care resources, but the resources invested might also be disproportional to their value, as there are growing concerns that DCIS may be overtreated.⁴

CONTEXT

• Key Objective

• To the best of our knowledge, our study is the first to evaluate the cost-effectiveness of all adjuvant treatment options for ductal carcinoma in situ (DCIS), focusing on clinically relevant patient subgroups.

• Knowledge Generated

• Radiation treatment (RT) is cost-effective in the majority of patients with standard-risk DCIS and in some younger patients with good-risk DCIS. Hormonal therapy is clinically suboptimal, whether alone or in combination with RT, as the side effects likely outweigh the benefits.

• Relevance

• Shared decision making with patients should be prioritized, as less medical therapy may lead to improved quality of life. Furthermore, there has been an increasing trend to replace RT with hormonal therapy for patients with good-risk DCIS; our findings suggest that such patients may be better served with observation or, if desired, RT alone.

Current treatment recommendations for DCIS include surgical resection with or without radiation treatment (RT) and hormonal treatment. The Early Breast Cancer Trialists' Collaborative Group (EBCTCG) meta-analysis found that RT after breast-conserving surgery (BCS) approximately halved ipsilateral breast tumor recurrence (IBTR) from 28% to 13% at 10 years, with no survival benefit.⁵ The UK/ANZ DCIS trial used a randomized 2 × 2 design, which was instructive in assessing the benefit of hormonal treatment: tamoxifen (TAM) alone had a smaller IBTR benefit than RT, but together with RT led to the lowest risk of ipsilateral recurrence, although still with no survival benefit.⁶ These results are consistent with the combined analysis of the NSABP B17/B24 trials⁷ and have led to the widely practiced strategy of dual adjuvant therapy in the United States.

In the absence of a survival improvement, the benefit of these treatments might be outweighed by their costs, in terms of both health care resources and adverse effects. As such, recent efforts have been aimed at identifying patients at low risk of recurrence for which adjuvant treatment may be omitted. The RTOG 9804 trial was a randomized trial evaluating the use of RT in patients with good-risk DCIS, defined as ≤ 2.5 cm, low-to-intermediate grade, and mammographically detected lesions with final margins ≥ 3 mm. Although RT reduced IBTR by 75%, the 10-year IBTR was < 10% with observation alone, and the trial concluded that the use of RT should be decided together with the patient.⁸ In practice, many of these good-risk patients who forego RT are instead recommended to receive hormonal treatment,⁹ without any clear evidence to support this strategy. Patient preferences also vary widely,^10,11 with only 30% of patients completing a full course of hormonal therapy.¹²

Cost-effectiveness is a useful measure of a treatment's value, synthesizing costs, efficacy, patient preferences, and quality of life (QOL) into a single comparative metric. The oncology community is particularly interested in questions of whether the cost and adverse effects of RT outweigh the burden of future recurrences and similarly whether five years of hormonal treatment is worth it for patients with DCIS.¹³ We therefore conducted a cost-effectiveness analysis that evaluates all potential adjuvant treatment strategies for DCIS after breast-conserving surgery across a cohort of patients varying in age and level of risk.

METHODS

Model Overview

We constructed a Markov model to simulate a broadly representative cohort of patients diagnosed with DCIS who underwent BCS followed by one of the six adjuvant treatment strategies (Fig 1). The patient cohort was modeled using annual transitions between health states, each with associated costs and QOL effects (known as utilities). Treatment strategies included the following: observation, TAM alone, aromatase inhibitor (AI) alone, RT alone, RT + TAM, or RT + AI. Patients who were 40 years old were not eligible for the AI-inclusive strategies. For patients receiving TAM or AI, compliance was modeled to drop off annually, with 70% of patients completing a standard 5-year course, as reported in the NSABP B35 and IBIS-II DCIS trials.^14,15 Within the RT arms, hypofractionated whole breast radiation was used based on clinical guideline recommendations and evidence supporting its equivalence with conventional fractionation.^16-19 After treatment, all patients initially entered a disease-free state, from which they could develop an IBTR or contralateral breast cancer (CBC) of DCIS or invasive histology. Patients who developed an invasive IBTR or CBC were at increased risk of breast cancer–specific death (through intervening metastases), whereas those with a DCIS recurrence were not, consistent with long-term NSABP B17/B24 trial results.⁷ All patients were at risk of all-cause death per actuarial life tables.²⁰

FIG 1.

Model diagram. After undergoing BCS for DCIS, patients were treated with one of the six adjuvant treatment strategies: observation, TAM alone, AI alone, RT alone, RT + TAM, or RT + AI. From the disease-free state, patients were at annual risk of an ipsilateral or contralateral recurrence of DCIS or invasive histology. Those who developed invasive recurrences were at risk of metastases and breast cancer–specific death, whereas all women were at risk of death because of other causes. AI, aromatase inhibitor; BCS, breast-conserving surgery; DCIS, ductal carcinoma in situ; RT, radiation treatment; TAM, tamoxifen.

Clinical Data Sources

Model inputs are given in Table 1. Baseline annual risk of all recurrence types was abstracted from the observation arm of the NSABP B17 trial for standard-risk DCIS.⁷ For good-risk DCIS, baseline IBTR risks were derived from the RTOG 9804 trial,⁸ adjusting for the use of TAM in the observation arm, and converted to annual transition probabilities (Data Supplement, online only), with no change in CBC risks. For TAM alone, we used a meta-analysis of the NSABP B24 and UK/ANZ DCIS trials to adjust baseline recurrence risks.²¹ For AI alone, we used a meta-analysis of the NSABP B35 and IBIS-II DCIS trials, both of which compared AI with TAM, to further adjust recurrence risks.²² For RT alone, the EBCTCG meta-analysis of four randomized RT trials was used to adjust IBTR risks⁵; a separate EBCTCG meta-analysis was used to model a possible increase in the risk of CBC after RT.²³ Although the finding was nonsignificant, we chose to be conservative in understanding all potential risks involved with RT. For the combined arms of RT + TAM and RT + AI, we used a combination of the above sources to derive all relevant risk ratios. The relative risk of AI versus TAM was equivalently modeled with or without RT, consistent with the IBIS-II DCIS trial.¹⁵ There were no IBTRs modeled after 15 years, as the risk of recurrence is low after this time interval⁶; however, the risk of CBC continued until age 85 years, with both the protective effect of TAM/AI and the deleterious effect of RT diminishing 15 years after treatment.^23,24

TABLE 1.

Model Parameters

Baseline annual risks of all major adverse events (Data Supplement) were derived from the placebo arm of the NSABP P-1 trial.²⁵ We then used meta-analyses of large randomized controlled trials to adjust these risks for each treatment strategy (Data Supplement), including relative protective benefits of TAM on bone fractures and AI on endometrial cancers and thromboembolic events.^26,27 For heart disease attributable to RT, we derived long-term risks from Darby et al.²⁸ Adverse events were modeled temporally: TAM/AI risks ended after treatment completion,²⁴ whereas RT risks of heart disease and secondary lung malignancy continued perpetually.^23,28

Salvage treatment was modeled using a combination of guideline-recommended care and observational data, accounting for variations in patterns of care. If patients were on TAM/AI at the time of recurrence (within the first 5 years), hormone therapy was discontinued given the presupposition of hormone resistance. Upon IBTR, all patients who previously received RT underwent mastectomy, whereas for those who were RT-naive, 50% underwent BCS/RT and 50% underwent mastectomy.²⁹ Rates of mastectomy, BCS/RT, and BCS alone were separately modeled for CBCs of DCIS and invasive histologies, without bias for previous RT.^30,31

Cost and Utility Inputs

We abstracted costs and utilities from the published literature. Medicare reimbursement rates were used for all initial and salvage treatment (Table 1) and treatment of adverse events (Data Supplement).^32,33 Average generic wholesale acquisition costs were used for TAM and AI.³⁴ For AI-inclusive strategies, costs of biennial osteoporosis screening and selective generic bisphosphonate therapy for 25% of patients were included.^34-36 Societal patient time and transportation costs were derived from the literature.³⁷ Medicare expenditures for metastatic care, breast cancer–related death, and background death were also included.^38,39 All costs were adjusted to 2018 US dollars using consumer price index data for medical care.⁴⁰

Primary health state utilities were abstracted from a study that collected patient preference weights for disease-free and recurrence states for DCIS (Table 1).⁴¹ Utility multipliers were then applied for breast RT and hormonal treatment⁴²; patients only experienced treatment disutility as long as treatment lasted. We assumed that TAM and AI have comparable utility given equivalent rates of noncompliance and discontinuation.^14,15 Utility decrements were also applied for major adverse events (Data Supplement), salvage treatment, and metastatic progression.⁴³ Utilities were age-adjusted in 10-year increments.⁴⁴

Model Analysis

Using guidelines recommended by the Second Panel on Cost-Effectiveness in Medicine,⁴⁵ we performed cost-effectiveness analysis through both societal (base-case) and payer (Medicare) perspectives. Four patient subgroups were considered: age 60, standard-risk; age 60, good-risk; age 40, standard-risk; and age 40, good-risk. We used a lifetime horizon with annual cycle length and discounted all future costs and utilities by 3% annually. Utilities were accrued through each simulated year, resulting in quality-adjusted life-years (QALYs) as the measure of effectiveness. Strategies were rank-ordered by cost as per convention. A dominated strategy was the one that had higher costs and fewer QALYs compared with alternative strategies. Incremental cost-effectiveness ratios (ICERs), defined as the ratio of incremental costs to incremental QALYs, were calculated against the next costliest, undominated strategy. A strategy was considered to be cost-effective if the ICER was below a willingness-to-pay (WTP) threshold of $100,000 USD/QALY. For simplicity with multiple competing strategies, we also used net monetary benefit (defined as the difference between the monetary value of QALYs [QALYs × WTP] and cost) to select the optimal strategy. All analyses were conducted in TreeAge Pro 2020 (Williamstown, MA). The model was validated by comparing ten-year model outcomes with ten-year cumulative risk from the NSABP B17 trial observation arm risk-adjusted by treatment strategy (Data Supplement).

Sensitivity Analysis

We conducted comprehensive sensitivity analysis for each subgroup. Deterministic one-way sensitivity analyses were performed to assess the robustness of optimal strategy selection to variation in input parameters. We were also specifically interested in understanding the changes in cost-effectiveness in two populations: first, in endocrine receptor–positive patients, which we modeled using an alternate set of TAM risk ratios obtained from a secondary analysis of the NSABP B24 trial,⁴⁶ and second, in a real-world population with a pragmatic hormonal therapy compliance of 35%, which we assumed to have proportionally increased risks compared with the aforementioned trials, but with less effect on cost and QOL compared with the base-case. We conducted an additional sensitivity analysis of conventionally fractionated RT by varying direct and indirect costs. Finally, we performed probabilistic sensitivity analyses by simultaneously varying inputs across their given probability distributions using 100,000 Monte Carlo simulations. We used log-normal distributions for relative risks, gamma distributions for costs, and beta distributions for utilities.⁴⁷ Acceptability curves were created depicting each strategy's overall probability of cost-effectiveness at different WTP thresholds.

RESULTS

The results of the base-case analysis are presented in Table 2 and on a cost-effectiveness plane in Figure 2. RT alone was cost-effective for standard-risk patients, and observation was cost-effective for good-risk patients, across both age groups. RT alone had substantially more QALYs than observation for standard-risk patients, especially in the age 40 cohort (16.69 v 16.48), whereas the differences were less pronounced for good-risk patients. All other strategies were dominated. The results from a health care perspective are presented in the Data Supplement; there were no differences in optimal strategy for each subgroup.

TABLE 2.

Cost-Effectiveness Results From a Societal Perspective (Base-Case)

FIG 2.

Cost-effectiveness planes depicting QALYs versus cost for each patient subgroup: (A) age 60 years, standard-risk; (B) age 60 years, good-risk; (C) age 40 years, standard-risk; and (D) age 40 years, good-risk. Line-connected strategies are undominated, for which an incremental cost-effectiveness ratio is calculated (refer to Table 2). Those to the left of the line are dominated, as they have fewer QALYs and higher costs than alternative strategies. AI, aromatase inhibitor; QALY, quality-adjusted life-year; RT, radiation treatment; TAM, tamoxifen; USD, US dollars.

Strikingly, strategies including AI or TAM had fewer QALYs than observation in all patient subgroups. This can be explained by a large decrease in QOL during 5 years of hormonal treatment in exchange for only a small benefit in risk reduction (which is more pronounced in the contralateral breast, for which there is smaller absolute risk). This is despite the fact that costs for generic AI and TAM are marginal and even a small reduction in risk saves relatively large costs of salvage treatment for each recurrence prevented, thereby leading to lower overall costs variably across subgroups.

In one-way sensitivity analyses, observation became cost-effective for older, standard-risk patients within the range of several RT input variables, as depicted in Figure 3A: an RT cost of $11,071 USD or higher, an RT utility of 0.82 or lower, or an invasive IBTR RT risk reduction of 0.55 or higher. Changes in variables for any other treatment did not shift the optimal strategy for these patients. Similarly, for younger, good-risk patients, RT alone became cost-effective with changes in specific RT variables, as depicted in Figure 3B: an RT utility of 0.95 or higher, an RT relative risk of invasive CBC of 1.05 or lower, an RT cost of $6,241 USD or lower, or an invasive IBTR RT risk reduction of 0.39 or lower. The optimal strategies for older, good-risk patients and younger, standard-risk patients were robust to any model input changes.

FIG 3.

Tornado diagrams showing a collection of one-way sensitivity analyses, arranged in descending order of potential uncertainty within the model. (A) For age 60 years, standard-risk patients, the base-case ICER for RT alone is $68,393 USD/QALY and observation becomes cost-effective (ICER > WTP) at an RT cost of $11,071 USD or higher, an RT utility of 0.82 or lower, or an Inv-IBTR RT risk reduction of 0.55 or higher. (B) For age 40 years, good-risk patients, the base-case ICER for RT alone is $134,039 USD; RT alone becomes cost-effective (ICER < WTP) at an RT utility of 0.95 or higher, an RT relative risk of Inv-CBC of 1.05 or lower, an RT cost of $6,241 USD or lower, or an Inv-IBTR RT risk reduction of 0.39 or lower. CBC, contralateral breast cancer; DCIS, ductal carcinoma in situ; IBTR, ipsilateral breast tumor recurrence; ICER, incremental cost-effectiveness ratio; Inv, invasive; QALY, quality-adjusted life-year; RR, relative risk; RT, radiation treatment; USD, US dollars; WTP, willingness-to-pay.

In the sensitivity analysis of a modeled endocrine receptor–positive population, there were small decreases in costs and small increases in QALYs for the AI- and TAM-alone strategies compared with the base-case; however, there was no change in optimal strategies (Data Supplement). Similarly, in the sensitivity analysis of real-world hormonal therapy compliance, there were modest increases in QALYs compared with the base-case, but the optimal strategies for each subgroup remained unchanged (Data Supplement). In the sensitivity analysis of conventionally fractionated RT, RT alone became marginally cost-ineffective for older, standard-risk patients, with an ICER of $106,315 USD/QALY (Data Supplement). As expected, RT strategies were less cost-effective across all subgroups than they were in the hypofractionated setting.

The probabilistic sensitivity analyses results are depicted in Figure 4. At the base-case WTP threshold of $100,000 USD/QALY, the probability that RT alone was cost-effective for older and younger standard-risk patients was 76% and 98%, respectively, whereas the probability that observation was cost-effective for older and younger good-risk patients was 97% and 66%, respectively. At WTP thresholds of $50,000 USD/QALY and $150,000 USD/QALY, the notable differences were that observation was cost-effective in 67% of simulations for older, standard-risk patients and RT alone was cost-effective in 56% of simulations for younger, good-risk patients, respectively.

FIG 4.

Acceptability curves generated from probabilistic sensitivity analyses for each patient subgroup: (A) age 60 years, standard-risk; (B) age 60 years, good-risk; (C) age 40 years, standard-risk; and (D) age 40 years, good-risk. AI, aromatase inhibitor; QALY, quality-adjusted life-year; RT, radiation treatment; TAM, tamoxifen; USD, US dollars; WTP, willingness-to-pay.

DISCUSSION

The management of DCIS presents a prime opportunity to find and reward value in medicine, because of its wide prevalence, excellent prognosis, and array of viable treatment approaches, each with its own cost and side effect profile. Although previous studies have looked at the cost-effectiveness of RT,^33,48 to the best of our knowledge, no study has evaluated hormonal therapy, alone or in combination with RT, or stratified by age or risk, all of which are important factors in the decision-making process for patients with DCIS.

We found that for patients of any age with standard-risk disease, RT alone was cost-effective, whereas for patients with good-risk disease, observation was cost-effective. Observation may also be an optimal choice for older, standard-risk patients who are especially keen to avoid RT or who may potentially experience a high burden of treatment. It appears that there are a variety of situations for which RT may be cost-effective for younger, good-risk patients. We used the EBCTCG-derived RT risk reduction of 0.46 for IBTR⁵ across all subgroups (only the baseline annual event probabilities were adjusted for good-risk patients), but the RTOG 9804 trial found RT to have a risk reduction of 0.25,⁸ which is below the threshold of 0.39 for which RT alone would be cost-effective. Second, if RT carries a relative risk of an invasive CBC of 1.05 or less, it would be cost-effective. We conservatively modeled this at 1.14 using the EBTCG meta-analyses, although they did not find it to be statistically significant; it is certainly plausible that RT carries less than a 5% increased relative risk of contralateral breast secondary malignancy with modern techniques.²³ Finally, younger patients are more likely to seek RT, because of both increased anxiety about a future recurrence and less hardship in completing treatment,⁴⁹ and therefore may have higher RT utility. In summary, our findings align with the conclusions of the RTOG 9804 trial, in that shared decision making with patients is central to the choice to use or forego RT.⁸

Our results are unique in that they suggest no role for hormonal therapy in the adjuvant treatment of DCIS. Hormonal therapy can decrease the total cost of care by reducing expensive salvage therapies, but the QOL detriment that patients experience for 5 years far outweighs the modest clinical benefit on IBTR reduction. In summary, the trade-off between efficacy and side effects is not favorable for hormonal treatment, either alone or in combination with RT. On the other hand, hormonal treatment provides a protective effect on the contralateral breast, which is greater in magnitude than its effect on ipsilateral recurrences.^6,7 In fact, the UK/ANZ trial concluded that the use of TAM should primarily be reserved for the prevention of new contralateral disease.⁶ This implies that hormones should be used selectively for patients who may be at high risk of de novo breast cancer on the basis of traditional risk factors, rather than for all patients with DCIS or as a substitute for RT.

The poor QOL that patients experience while on hormonal therapy is evidenced by their poor tolerability, leading to only 30% compliance with a full 5-year course.^12,50 In our sensitivity analysis of real-world hormonal therapy compliance, QOL was improved, yet still inferior to observation across all subgroups. Physicians also have markedly different practice patterns regarding the use of hormones; in one survey, 56% of North American respondents indicated that they always use TAM for DCIS, whereas only 22% respondents from Europe did.⁵¹ Therefore, to some extent, patients and physicians are already practicing omission of hormonal therapy, but our findings serve as objective data to validate these patterns of care.

One of the National Cancer Institute Breast Cancer Steering Committee's strategic priorities is to decrease toxicity and costs of treatments with minimal clinical benefit.⁵² Included within this priority is an aim to decrease the use of RT, and accordingly, there has been a great deal of interest in omitting RT for both DCIS and early-stage invasive breast cancer.⁵³ Our results should also motivate interest and prospective exploration of omitting hormonal therapy from routine DCIS management.

An important limitation is that we modeled hypofractionated RT to have equal efficacy to conventionally fractionated RT, on the basis of currently available evidence. Although we await the results of the TROG 07.01 trial, the definitive trial of hypofractionated RT for DCIS,⁵⁴ the recently published DBCG HYPO trial included patients with DCIS and found that long-term efficacy and toxicity were similar or improved in the hypofractionated arm, potentially strengthening our results.⁵⁵ Furthermore, the degree to which poor compliance affects recurrence risk is unknown; we assumed a proportional increase in risk, and our findings aligned with the base-case in suggesting that less hormonal therapy leads to better QOL. An additional limitation is that our model is a simplification of the clinical course of DCIS as we did not model the potential for multiple local and/or contralateral recurrences, or local recurrences after 15 years, both of which are relatively rare.^6,56 Finally, utility is a general measure of QOL and, as in all cost-effectiveness analyses, may vary widely between patients.

In conclusion, we found that RT alone is the optimal postoperative strategy for patients with standard-risk DCIS. Observation is cost-effective for patients with good-risk DCIS, but there are a variety of situations in which RT may be cost-effective for younger patients. Finally, hormonal therapy is likely a suboptimal choice for most patients with DCIS, either alone or in combination with RT, as it appears that the negative impact on QOL outweighs the benefit in recurrence risk reduction.

AUTHOR CONTRIBUTIONS

Conception and design: Apar Gupta, Sachin R. Jhawar, Mutlay Sayan, Zeinab A. Yehia, Bruce G. Haffty, Shi-Yi Wang

Financial support: Bruce G. Haffty

Administrative support: Bruce G. Haffty

Provision of study materials or patients: Bruce G. Haffty

Collection and assembly of data: Apar Gupta, Bruce G. Haffty

Data analysis and interpretation: Apar Gupta, Mutlay Sayan, James B. Yu, Shi-Yi Wang

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Cost-Effectiveness of Adjuvant Treatment for Ductal Carcinoma In Situ

The following represents disclosure information provided by the authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).