FormalPara Key Summary Points

Why carry out the study?

To our knowledge, no previous studies have assessed the population-level impacts of second-generation HER2-targeted therapies [pertuzumab and ado-trastuzumab emtansine (T-DM1)] on the number of women who experience recurrence after treatment for HER2-positive early breast cancer (EBC).

The objective of this study was to assess the impact of HER2-positive treatment landscape changes following the introduction of pertuzumab and T-DM1 on cumulative population-level recurrences avoided since 2013 in the United States (US).

Importantly, our analysis aimed to estimate the actual population impact of these therapies (based on observed and forecasted utilization), rather than the potential population impact, which would be estimated by assuming full utilization by all eligible women.

What was learned from the study?

In steady-state equilibrium, the model estimated that real-world utilization of pertuzumab and T-DM1 will reduce the population-level number of recurrences by approximately 32%: in 2031, we estimate that, based on current utilization rates, 7226 women will experience recurrence, and fewer recurrences with neoadjuvant pertuzumab, continuation of pertuzumab in the adjuvant setting, and T-DM1 in the adjuvant setting in women with residual disease after neoadjuvant treatment are expected.

Given the improvement of HER2-targeted treatments, alongside increases in BC disease burden, we expect that the population-level impact of HER2-targeted treatments will accelerate over the next decade.

Our results suggest that utilization of HER2-targeted treatments in the US has the potential to change the epidemiology of HER2-positive EBC by preventing a substantial number of women from experiencing disease recurrence; these improvements may help inform on the future disease and economic burden of HER2-positive BC in the US.

Introduction

Breast cancer (BC) is the most common cancer among women in the United States (US). In 2022, an estimated 287,850 women were diagnosed with BC and 43,250 died from their disease [1]. The human epidermal growth factor receptor 2 (HER2) protein is overexpressed, or the HER2 gene is amplified, in approximately 15–20% of BCs [2], which increases the aggressiveness of the disease [3, 4]. The prognosis for HER2-positive early BC (EBC) was significantly improved following the introduction of the HER2-targeted monoclonal antibody trastuzumab, which was approved by the US Food and Drug Administration (FDA) in 2006 for the adjuvant treatment of nonmetastatic HER2-positive, node-positive BC [5].

Together, the NSABP B-31/NCCTG N9831 joint analysis [6], and the BCIRG-006 [7] and HERA [8] trials, demonstrated a substantial reduction in risk of recurrence with trastuzumab versus chemotherapy alone as adjuvant treatment in women with HER2-positive EBC. Despite these improvements, patients treated with adjuvant trastuzumab still faced a non-trivial risk of recurrence in the long term [9, 10].

Subsequent research demonstrated the importance of total pathologic complete response (t)pCR in treating EBC. tpCR refers to absence of residual invasive cancer in the breast and axillary lymph nodes, whereas pCR historically referred to absence in the breast only. Thus, the importance of neoadjuvant treatment was also demonstrated, as it is given prior to surgery, in patients with high-risk HER2-positive EBC. Specifically, patients with HER2-positive EBC who are treated in a neoadjuvant setting may achieve (t)pCR following HER2-targeted therapy in combination with chemotherapy, which lowers subsequent risk of recurrence or death relative to patients with residual invasive disease (RD) at surgery [11,12,13]. However, the CTNeoBC pooled analysis made clear that, while tpCR reduced the risk of recurrence, it did not eliminate it. As such, patients treated with trastuzumab in the neoadjuvant setting continued to experience a risk of recurrence, with this risk being greater for patients with RD [13].

Following on from trastuzumab, a second-generation HER2-targeted therapy was developed (pertuzumab), along with an antibody–drug conjugate (ado-trastuzumab emtansine; T-DM1). Both of these therapies have been evaluated as options for reducing the risk of recurrence following treatment of EBC. Pertuzumab is a monoclonal antibody that provides a more comprehensive signaling blockade when combined with trastuzumab. Its efficacy in EBC was first assessed in the neoadjuvant setting in the registrational NeoSphere study [14]. NeoSphere demonstrated that pertuzumab, trastuzumab, and chemotherapy improved the probability of achieving a pCR and a tpCR compared with trastuzumab and chemotherapy alone. In the adjuvant setting, the APHINITY study demonstrated increased invasive disease-free survival with pertuzumab, trastuzumab, and chemotherapy versus trastuzumab and chemotherapy [15, 16]. Pertuzumab received accelerated approval by the FDA in 2013 for use in combination with trastuzumab and chemotherapy as neoadjuvant treatment for HER2-positive locally advanced, inflammatory, or early-stage BC (tumors either greater than 2 cm in diameter, or node-positive) as part of a complete treatment regimen for EBC [17]. In 2017, this was converted to full approval and the indication was expanded for its use in combination with trastuzumab and chemotherapy as adjuvant treatment for HER2-positive EBC at high risk of recurrence. This was based on the results from the pivotal APHINITY study [15].

T-DM1 is an antibody–drug conjugate that combines trastuzumab with a cytotoxic agent, emtansine. In the KATHERINE trial, T-DM1 reduced disease recurrence compared with trastuzumab in patients with RD after neoadjuvant HER2-targeted therapy [18]. T-DM1 was approved by the FDA in 2019 for patients who had RD after neoadjuvant trastuzumab-based treatment [19].

Together, pertuzumab and T-DM1 radically altered the treatment landscape for patients with HER2-positive EBC. The current neoadjuvant standard of care for patients with high-risk HER2-positive EBC is pertuzumab plus trastuzumab [20, 21]. The current standard of care for patients with RD after neoadjuvant treatment is T-DM1 [20]. In addition, pertuzumab is widely used (in combination with trastuzumab) in both the adjuvant and neoadjuvant–adjuvant continuation settings.

The population-level impacts of trastuzumab on the number of women who experienced recurrence have previously been estimated [22]; however, to our knowledge, no studies have performed this type of analysis for second-generation HER2-targeted therapies (pertuzumab and T-DM1). The objective of this study was therefore to assess the impact of HER2-positive treatment landscape changes following the introduction of pertuzumab and T-DM1 on cumulative population-level recurrences avoided since 2013. Importantly, our analysis aimed to estimate the actual population impact of these therapies (based on observed and forecasted utilization), rather than the potential population impact, which would be estimated by assuming full utilization by all eligible patients.

Methods

Overview

We constructed a multi-year epidemiologic population treatment impact model using Excel (Microsoft, Redmond, WA, USA). The purpose of this model was to estimate annual recurrences for women treated for HER2-positive EBC (defined as locally advanced, inflammatory, or early-stage BC, with treatment given as part of a complete treatment regimen for EBC) between 2013 (the first approval of either molecule in the US) and 2031 (10 years from the conduct of this analysis). The flow of the model is presented in Fig. 1. Specifically, we developed annual estimates from 2013 to 2031 for the following parameters: (1) incidence of women with BC; (2) proportion of BCs diagnosed at stage I–III; (3) proportion of those women with HER2-positive disease; (4) treatment proportions for neoadjuvant-only, adjuvant-only, and neoadjuvant–adjuvant continuation treatment; and (5) therapeutic agent proportions within each of those treatment settings (i.e., chemotherapy only, trastuzumab ± chemotherapy, pertuzumab with trastuzumab ± chemotherapy, or T-DM1). The combination of these inputs led to model-generated annual synthetic cohorts. These cohorts were then followed over time to estimate the study’s primary endpoint (cumulative recurrences). This quantity was estimated by further incorporating extrapolated clinical trial data for each regimen of interest into the model.

Fig. 1
figure 1

Flow of the model. BC breast cancer, Chemo chemotherapy, EBC early breast cancer, H trastuzumab, P pertuzumab, RD residual disease, T taxane, T-DM1 ado-trastuzumab emtansine, tpCR total pathologic complete response. aOccult primary tumors are the only clinical exclusion among women with stage I–III disease prior to decision for preoperative (neoadjuvant) systemic therapy bRecurrences are inclusive of locoregional and contralateral recurrences

Full size image

Key input parameters are listed in Table 1. The historic and projected incidence of HER2-positive BC were estimated by combining data from several sources. We obtained US Cancer Statistics (USCS) data, which combine cancer registry data from the Centers for Disease Control and Prevention’s National Program of Cancer Registries (NPCR) and the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program [23, 24]. This dataset includes cancer incidence from central cancer registries covering all 50 states, the District of Columbia, and Puerto Rico [25]. We included new cases of primary invasive BC (International Classification of Diseases for Oncology, Third Edition code C50) among diagnosed women from 2001 to 2017 [26,27,28]. For cancer incidence projections in the years from 2018 onward, we assumed changes to the underlying US population structure based on projections from the United Nations [29]. We assumed no secular increase in BC incidence rates apart from changing demographics. With these assumptions, our model projected an estimated annual growth in BC case counts of less than 1% from 2018 onward. For the proportion of women diagnosed with HER2-positive disease and stage I–III disease, we separately reviewed annual estimates of these parameters from 2010 to 2018 in the USCS database. Because there was little variation over time, we used the stabilized mean estimates for each parameter across all years [25]. The proportion of women receiving adjuvant treatment was determined based on 10-year stabilized estimates from real-world data sources [24, 30]. The proportion of women receiving neoadjuvant treatment was determined based on annual proportions from real-world data sources through 2019 [30] and internal estimates for years 2020 onward (Genentech, Inc., data on file).

Table 1 Key model input parameters
Full size table

Allocation of women to treatment combinations within each of the four settings was estimated using market share data. In the base case, incident patients between 2013 and 2021 were assumed to be treated according to actual data on real-world utilization. For example, market data (Genentech, Inc., data on file) indicated that, in 2021, 85% of women treated in the neoadjuvant setting received pertuzumab, trastuzumab, and chemotherapy; 12% received trastuzumab plus chemotherapy; and 3% received chemotherapy alone. The base case model thus replicated this allocation to treatment regimens. For future years (2022–2031), the model fixed market shares at 2021 values. This base case estimate was then compared with a series of counterfactual scenarios (described below) in which pertuzumab, T-DM1, or both were assumed to be unavailable in one or more treatment settings.

In these counterfactual scenarios, women were proportionally reassigned to the remaining regimens. For example, in the base case scenario described above, 12% of women in the neoadjuvant setting were treated with trastuzumab plus chemotherapy and 3% were treated with chemotherapy alone. Thus, for the counterfactual in which pertuzumab is removed as a treatment option, these proportions were set to 80% and 20%, respectively.

Both the setting-specific and aggregate impact of pertuzumab and T-DM1 on BC recurrences were estimated by calculating the differences in modeled recurrences between the base case and counterfactual scenarios, or between pairs of counterfactual scenarios.

To estimate the yearly number of recurrences per treatment setting and each available treatment option, invasive disease-free survival (IDFS) curves from various clinical trials and disease settings [7, 8, 15, 18, 32, 37,38,39,40,41,42,43] were extrapolated over a 10-year time horizon. The extrapolations were adjusted for cure proportion, duration of treatment effect, and background risk of death. The results and methodology used to extrapolate the IDFS data in the neoadjuvant setting were the same as those presented by Sussell et al. [33] and are further described there. For women treated in the adjuvant-only setting, the methodology was aligned with that presented by Garrison et al. [35] but updated using the latest clinical trial data [36], with some further modification in the parametric model and assumptions.

Depending on the neoadjuvant and adjuvant treatment algorithm, the resulting annual transition probability of having a disease recurrence was then estimated by first finding the yearly proportion of women with an event, as informed by the 10-year extrapolated IDFS curves. Second, we removed the proportion of events expected to be deaths (without prior recurrence), estimated based on the proportion of first IDFS events that were reported as death, as observed from clinical trials [18, 36]. For more details about the modeled survival inputs and assumptions, see Table 1.

The cohort probability of developing a recurrence is further impacted by two additional model inputs: the probability of tpCR and adjuvant market share assumptions.

Outcomes

We estimated annual and cumulative disease recurrences between 2013 and 2031 under four scenarios: Scenario A, base case scenario: market uptake of T-DM1 for RD and pertuzumab for remaining settings (i.e., neoadjuvant, adjuvant continuation, and adjuvant-only therapy) through 2021; projected utilization rates were fixed at 2021 levels through 2031. Scenario B, primary counterfactual: no women received pertuzumab or T-DM1 in any setting; instead, women were treated either with the previous standard of care (trastuzumab plus chemotherapy) or with chemotherapy alone. Scenario C, secondary counterfactual: the same as the primary counterfactual, but women received pertuzumab in the neoadjuvant setting only at historical/projected utilization rates. Scenario D, tertiary counterfactual: the same as the secondary counterfactual, but also women who achieved pCR received adjuvant continuation of pertuzumab at historic/projected market share rates, and women receiving adjuvant-only therapy received pertuzumab at historic/projected market share rates. Comparison of the results of the base case and counterfactual scenarios allows for estimation of the population-level impact of pertuzumab and T-DM1 on BC recurrence: the incremental impact of neoadjuvant pertuzumab was calculated as the difference in cumulative recurrences between the base case scenario and the secondary counterfactual; the incremental impact of adjuvant pertuzumab was calculated as the difference in cumulative recurrences between the secondary and tertiary counterfactuals; lastly, the incremental impact of adjuvant T-DM1 was calculated as the difference in cumulative recurrences between the tertiary counterfactual and the base case scenario. The aggregate population impact of pertuzumab and T-DM1 (across all settings) was estimated as the difference between the base case and the primary counterfactual. Intuitively, the aggregate impact of both therapies can also be constructed as the sum of the three setting-specific incremental impacts described above.

Sensitivity and Scenario Analyses

Probabilistic Sensitivity Analysis

We used probabilistic sampling (a Monte Carlo process) to generate confidence intervals for point estimates due to uncertainty in trial estimates of pCR and disease-free survival rates. We generated 1000 replicates of the analysis derived from repeatedly resampling the probability distributions described in Table 1. Confidence intervals for study outcomes were obtained using the 2.5th and 97.5th percentiles of the distributions of these replicates.

Scenario Analysis

In our base case parameterization, we assumed that the real-world regimen-specific tpCR rates were equivalent to those published in the NeoSphere trial [14]. We also conducted a scenario analysis in which we instead used the tpCR rates provided by a recently published pooled analysis of regimen-specific outcomes in the treatment of HER2-positive EBC [31]. There are advantages and disadvantages of each of these sources. NeoSphere randomized women to pertuzumab, trastuzumab, and chemotherapy and compared this arm with trastuzumab plus chemotherapy (as well as pertuzumab plus chemotherapy and pertuzumab plus trastuzumab alone), and thus had strong internal validity. However, NeoSphere included a relatively small number of women in each arm, resulting in relatively large confidence intervals. NeoSphere also included only four cycles of dual HER2-targeted therapy in the neoadjuvant setting. The pooled analysis, on the other hand, included data from several trials, reducing the overall level of statistical uncertainty. In addition, several of these trials contained chemotherapy backbones of duration and timing (neoadjuvant vs. after surgery) that may be more consistent with clinical practice in the US. However, the pooled analysis was not powered for comparisons between the treatment arms; therefore, the tpCR endpoint results were descriptive and did not include any statistical adjustments across arms; consequently, the tpCR rate estimates, when compared across treatment arms, were essentially naive averages.

Compliance with Ethics Guidelines

This modeling analysis was conducted using aggregate data from previously conducted trials and studies; no individual patient data were generated or accessed during the analysis. The public use registry data used to provide inputs for the epidemiology component of our model was accessed with full permissions for use and publication.

Results

Primary Analysis

Approximately 889,057 women were predicted to be diagnosed with stage I–III HER2-positive BC from 2006 to 2031 in the US and potentially indicated for a HER2-targeted treatment. Figure 2 shows the estimated annual recurrences in HER2-positive BC between 2013 and 2031 under the base case scenario and each of the counterfactual scenarios among these women. Each of the four time series is generally upward-trending, reflecting the growth in the size of the population at risk for BC. This is a function of both the increase in the number of women at risk for disease as a whole, as well as the projected increase in average age of the general US population over time. This latter aspect is relevant because BC risk increases with age. The impact of neoadjuvant pertuzumab on population-level BC recurrence is represented by the difference between the primary counterfactual (Scenario B), in which neither pertuzumab nor T-DM1 was available in any setting (black series) and the secondary counterfactual (Scenario C), in which neoadjuvant pertuzumab was available (red series with short dashes). The divergence between these series began shortly after commercial approval in 2013. The incremental impact of adjuvant pertuzumab is visualized as the difference between the secondary and tertiary counterfactuals, in which pertuzumab was additionally available in the adjuvant setting (gray series with long dashes; Scenario D). Divergence began following commercial approval of adjuvant pertuzumab in 2017. The incremental impact of adjuvant T-DM1 is visualized as the difference between the base case and tertiary counterfactuals (i.e., Scenarios A and D), in which T-DM1 is available in the adjuvant setting (yellow series with medium dashes). Divergence began following commercial approval of T-DM1 in 2019. The combined impact of pertuzumab and T-DM1 on avoided breast cancer recurrences is thus represented by the difference between the base case and primary counterfactual scenarios. In steady-state equilibrium (i.e., after accounting for uptake lags), the model estimated that real-world utilization of pertuzumab and T-DM1 will reduce the population-level number of women who will experience HER2-positive BC recurrence by approximately 32%. For example, in 2031, we estimate that, based on current utilization rates, 7226 women will experience recurrence. In the counterfactual scenario in which neither pertuzumab nor T-DM1 is available, and women are instead treated with chemotherapy ± trastuzumab, we estimate that an additional 3394 women would experience recurrence.

Fig. 2
figure 2

Annual estimated recurrences of HER2-positive breast cancer in the United States. Base case versus counterfactual scenarios with reduced treatment options. BC breast cancer, EBC early breast cancer, H trastuzumab, P pertuzumab, RD residual disease, T taxane, T-DM1 ado-trastuzumab emtansine

Full size image

Figure 3a shows recurrences avoided following utilization of pertuzumab in the neoadjuvant-only setting. Both series are increasing over time due to two factors. First, as previously mentioned, the population at risk for BC is anticipated to increase over time, as the US population simultaneously ages and grows larger. Additionally, there has been a substantial increase in the fraction of women with HER2-positive EBC treated neoadjuvantly. In 2012, the year prior to the approval based on NeoSphere, only 20% of patients received systemic therapy prior to surgery. At present, approximately 57% of all patients do (Genentech, Inc., data on file).

Fig. 3
figure 3

Estimated annual recurrences in HER2-positive breast cancer in the United States, shown according to treatment setting. a With/without neoadjuvant pertuzumab (women treated in the neoadjuvant-only setting; Scenarios C and B). b With/without adjuvant continuation of pertuzumab (women treated in the neoadjuvant-only setting; Scenarios D and C). c With/without adjuvant pertuzumab (women treated in the adjuvant-only setting; Scenarios D and C). d With/without T-DM1 for women with residual disease (patients treated in the neoadjuvant-only setting; Scenarios A and D). The series in Fig. 3a are not identical with those in Fig. 2 because they exclude women treated in the adjuvant-only setting. BC breast cancer, EBC early breast cancer, H trastuzumab, P pertuzumab, RD residual disease, T taxane, T-DM1 ado-trastuzumab emtansine

Full size image

Figure 3b shows recurrences avoided, due to pertuzumab in the adjuvant continuation setting only, among the women who achieved a tpCR. As before, both series are increasing over time due to the increase in the size of the population at risk, as well as the shift toward neoadjuvant treatment and away from adjuvant-only treatment.

Figure 3c shows recurrences avoided due to pertuzumab in the adjuvant-only setting. Both series are decreasing over time. Although the size of the population at risk increases over time, this is more than offset by the shift toward neoadjuvant treatment and away from adjuvant-only treatment.

Figure 3d shows recurrences avoided due to T-DM1 in the adjuvant setting only for women with RD. As mentioned earlier, series are increasing over time due to the previously mentioned increase in the size of the population at risk, and substantial shift toward neoadjuvant treatment and away from adjuvant-only treatment. Although not directly comparable due to design differences, our estimates of population-level recurrences in the residual disease population were qualitatively similar to those recently published Hendrix et al. [44].

Sensitivity and Scenario Analyses

Table 2 presents the results of the probabilistic sensitivity analysis and the scenario analysis. The third column displays setting-specific and aggregate estimates of recurrences avoided over the entire relevant time period, as estimated in the base case analysis compared with the relevant counterfactual scenarios. The fourth column presents the 95% confidence interval for that base case estimate. In all instances, these exclude zero, indicating that results are statistically significant after accounting for joint uncertainty across multiple input parameters. The fifth column contains results from the scenario analysis in which the tpCR rates from the pooled analysis [31] were used in place of those from the NeoSphere study [14]. The primary difference that resulted from this shift was a halving of the estimate of recurrences avoided due to the use of neoadjuvant pertuzumab, from 9183 to 4396. As discussed above, the pooled analysis contained several trials with chemotherapy regimens with duration/timing that differed from that of the control arm of NeoSphere, resulting in a higher tpCR rate (and thus fewer recurrences) in the scenario without pertuzumab. Scenario analyses using the pooled analysis estimate of tpCR rate in women treated with pertuzumab, trastuzumab, and chemotherapy (42.1%) showed qualitatively similar results to those of NeoSphere (39.3%), but the rate for women treated with trastuzumab plus chemotherapy was substantially higher (33.6% vs. 21.8%). As a result, the alternative parameterization of the counterfactual without the option of neoadjuvant pertuzumab resulted in more women achieving tpCR (relative to the base case counterfactual parameterized by NeoSphere). Because tpCR is an important prognostic factor for breast cancer recurrence, this ultimately led to 5943 fewer recurrences avoided, relative to the base case model.

Table 2 Sensitivity analysis: recurrences avoided by setting
Full size table

Discussion

Prior to the advent of HER2-targeted therapies, HER2-positive BC had one of the poorest prognoses among BC variants [45]. It is generally recognized that HER2-targeted therapy, including trastuzumab, pertuzumab, and T-DM1, has fundamentally changed the epidemiology of HER2-positive BC [46]. Specifically, women diagnosed with HER2-positive EBC may be eligible for HER2-targeted therapies in neoadjuvant and/or adjuvant settings, which may delay or prevent disease progression in the curative setting. Using an epidemiologic population treatment impact model of US women diagnosed with HER2-positive EBC from 2013 onward, we estimated that real-world uptake of pertuzumab and T-DM1 will reduce the population-level incidence of BC recurrence by nearly one third. These estimated reductions were driven by relative treatment benefits accruing to women in the long term, based on separation of survival curves demonstrated in clinical trials. Parameter uncertainty explored through probabilistic sensitivity analyses did not alter the interpretation of the results, suggesting a similar magnitude of impact from the introduction of HER2-targeted treatment. Because our model incorporated real-world data on the utilization of targeted therapies in different settings, our results should be viewed as a combination of two constructs: estimates of what the actual impact of targeted therapy in the US has been between 2013 and the present; and estimates of what the impact may be over the next decade, based on current utilization. This is fundamentally different from studies that assume that all eligible patients receive the therapies under consideration, which is an approach that is sometimes used in analyses sharing our general design [47]. Had we made a base case modeling assumption that all eligible women would receive the indicated treatments, our analysis would have yielded substantially larger estimates of recurrences avoided. This is primarily because many patients, for a variety of reasons, do not receive the recommended treatment according to treatment guidelines. For example, our data suggested that approximately half of patients treated with neoadjuvant pertuzumab, trastuzumab, and chemotherapy who achieved a tpCR discontinue pertuzumab, despite evidence that continuation of dual HER2-targeted therapy reduces the risk of breast cancer recurrence substantially [31].

Our analysis has important limitations. We assumed that real-world effectiveness would be equal to clinical trial efficacy. This assumption provides a particular threat to the validity of our estimates for subgroups not well represented in the clinical trials. Future work using real-world data could help to inform potential transportability of data from HER2-targeted clinical trials. Similarly, we did not consider potential differences in treatment utilization or treatment adherence among minority populations. For example, a recent study of elderly populations in the US demonstrated substantial disparities in trastuzumab utilization [48], underscoring the importance of equitable access to medicines to ensure that future population-level improvements are experienced optimally across population groups. Given the aging population of our study, additional work may help to inform specific treatment needs of elderly populations as well as those of different racial and ethnic groups.

Additionally, our study is subject to limitations inherent to epidemiologic prediction models. Specifically, the model estimates may not appropriately account for disease incidence or population structure changes over time or reflect actual treatment practice or changes that occurred after the model was developed. In sensitivity analyses, we observed that our findings are sensitive to differences in real-world treatment uptake, and we did not account for treatment discontinuation or lack of adherence to the labels. The differences between the tpCR rates observed in the NeoSphere trial [14] and in the pooled analysis [31] create an additional source of uncertainty, one which we attempted to address in our scenario analysis. The model also assumed that no more efficacious HER2-targeted agents would be introduced for EBC during the modeling period. This is fundamentally a conservative assumption, to the extent that if more efficacious targeted agents are introduced in the future, our impact estimates will be biased toward the low end of the impact spectrum. Neratinib is also a treatment option for patients with HER2-positive EBC; however, the market share of neratinib is quite low in the US (approximately in the order of 1% according to our data). Thus, adding neratinib as a treatment option would have added complexity without meaningfully impacting the results. Given this, we elected to exclude it from our analyses. Finally, the impact of the COVID-19 pandemic on staging and disease incidence was not accounted for, and we did not consider women who would have been treated off-label, either before or during our analysis period.

Conclusions

Given the improvement of HER2-targeted treatments, alongside increases in BC disease burden, we expect that the population-level impact of HER2-targeted treatments will accelerate over the next decade. Our results suggest that utilization of HER2-targeted treatments in the US has the potential to change the epidemiology of HER2-positive EBC by preventing a substantial number of women from experiencing disease recurrence. These improvements may help inform our understanding of the future disease and economic burden of HER2-positive BC in the US.