Introduction

Human epidermal growth factor receptor 2-positive (HER2+) breast cancer (BC) is a clinically and biologically heterogeneous disease [1] characterized by the amplification of the ERBB2/HER2 gene and/or overexpression of its related kinase receptor protein [2]. This tumor subtype comprises around 15–20% of all BCs and has been associated with a high risk of recurrence and poor outcomes [3, 4].

HER2-targeted agents have radically improved the prognosis of HER2+ early BC (EBC) offering the possibility to de-escalate standard chemotherapy in selected subgroups [5]. The neoadjuvant setting provides the best scenario for treatment de-escalation considering that patients achieving a pathological complete response (pCR) have favorable long-term outcomes in terms of disease-free survival and overall survival [6].

Several studies have investigated predictive factors of pCR to neoadjuvant treatment. Imaging tools that could guide the response to preoperative therapy are of great interest, mainly the potential utility of [18Fluorine]fluorodeoxyglucose ([18F]FDG)-positron emission tomography (PET). The association between early treatment response on ([18F]FDG-PET) and clinical outcomes has been evaluated in patients with HER2+ BC in metastatic and neoadjuvant settings [4, 7,8,9,10,11]. These studies demonstrated that early metabolic evaluation using [18F]FDG-PET might identify HER2+ tumors with high anti-HER2 sensitivity and an increased likelihood of achieving a pCR on neoadjuvant HER2 blockade [10,11,12].

PHERGain (NCT03161353) is an international, randomized, open-label, phase II trial that aims to assess the efficacy of a chemotherapy-free strategy based on a dual HER2 blockade with trastuzumab-pertuzumab (+ endocrine therapy for hormone receptor [HR]-positive tumors) as neoadjuvant and adjuvant therapy in HER2+ EBC patients through an [18F]FDG-PET and pCR-adapted strategy [5]. It is assessing whether [18F]FDG-PET along with the pathological response could identify tumors with high anti-HER2 sensitivity to avoid standard chemotherapy in subsequent cycles (neoadjuvant and adjuvant setting) in patients with an early [18F]FDG-PET response that achieved a pCR with exclusive dual HER2 blockade with trastuzumab-pertuzumab. For this purpose, breast lesions must be SUVmax ≥ 1.5 × SUVmean liver + 2 SD, an inclusion criterion that was not met in a significant number of patients (screening failure).

The clinical, pathological, and molecular characteristics of these screening failures need to be investigated in order to guide more adequate patient selection. Numerous studies have reported an association between tumor [18F]FDG uptake in [18F]FDG-PET and both molecular subtypes of BC, as well as clinicopathological characteristics [11, 13,14,15,16,17,18,19,20,21,22].

To investigate this, we designed the RESPONSE, a substudy of the PHERGain trial, in which we aimed to obtain information regarding the clinicopathological and molecular characteristics of tumors that can impact on the [18F]FDG-PET evaluation of HER2+ tumors, and, therefore, on its ability of prediction (diagnosis/response). Here, we report the results of this subanalysis, which included all patients with HER2+ EBC screened in the PHERGain trial.

Material and methods

See the full version of the “Material and methods” section as supplementary material (references [5, 23,24,25]).

Our substudy included patients from the PHERGain trial (NCT03161353) [5] that had previously been untreated, HER2+ , stage I-IIIA, invasive and operable BC, with ≥ 1 [18F]FDG-PET target breast lesions ≥ 1.5 cm by MRI or ultrasound at baseline. PET[+] criteria patients were those with SUVmax levels ≥ 2.5 based on the formula for breast lesions (SUVmax ≥ 1.5 × SUVmean liver + 2 standard deviation [SD]), and their matched PET[−] counterparts had SUVmax levels ≥ 2.5. A visual comparison of two typical patients of each cohort is shown in Supplementary Fig. 1.

The objectives of this substudy were as follows: (1) to evaluate the association between SUVmax and PET[+] and PET[−] criteria at baseline with the clinicopathological features in all screened patients with a tumor size > 1.5 cm by MRI (n = 500) and (2) to analyze differences in sTILs and gene expression using PAM50 (intrinsic subtyping and ROR scores) and Vantage 3D™ Cancer Metabolism Panel in a matched cohort of 21 PET[−] and 21 PET[+] patients.

This study was performed in accordance with guidelines of the International Conference on Harmonization and ethical principles outlined in the Declaration of Helsinki. Written informed consent was required before enrolment, and all participants agreed to study-specific procedures. Approvals from the following regulatory authorities and ethics committees were obtained: Comité Ético de la Investigación con Medicamentos del Hospital Arnau de Vilanova (Spain), Comité de Protection des Personnes EST-III Hôpital de Brabois (France), Ethikkommission der Medizinischen Fakultät Heidelberg (Germany), The Ethics Committee of the Institut Jules Bordet (Belgium), CEIC—Comissão de Etica para a Investigação Clinica, Parque Saude Lisboa (Portugal), Fulham Research Ethics Committee Charing Cross Hospital (United Kingdom), and Comitato Etico ASL Brindisi, Comitato Etico della provincia Monza Brianza, Comitato Etico Istituto Europeo di Oncologia e Centro Cardiologico Monzino, Comitato Etico Val Padana, Comitato Etico di Area Vasta Emilia Centro, and Comitato Etico dell’Area Vasta Emilia Nord (Italy).

Statistical analysis

In all screened patients of PHERGain trial with a tumor size > 1.5 cm by MRI, unadjusted and adjusted analyses were performed to assess the relationship between SUVmax at baseline and [18F]FDG-PET status ([−] or [+]) with clinicopathological characteristics (age, tumor stage, tumor size, nodal status, HR status, HER2 immunohistochemistry status, carcinoma type, tumor grade, and Ki67 proliferation index).

Subsequently, we randomly selected 21 PET [−] criteria patients with SUVmax levels < 2.5 among 75 PET [−] criteria patients screened. We matched them with 21 patients with SUVmax levels ≥ 2.5 (PET[+] patients) based on patient characteristics previously associated with [18F]FDG-PET status (tumor size, nodal involvement, and histological grade). PET[−] and PET[+] patients had the same tumor stage, nodal status, and histological grade in the matched analysis. PAM50 intrinsic subtyping, ROR scores, and cancer metabolism gene expression were compared according to [18F]FDG-PET status in the matched cohorts.

For all statistical analyses, p-value < 0.05 was considered statistically significant. Multiple testing issues with gene expression were controlled with a false discovery rate using a threshold of q-value < 5%. Results from overall correlative analyses should be considered descriptive because of the small number of samples and unadjusted sequential testing.

Results

Patients

All patients screened in the PHERGain trial with a tumor size > 1.5 cm by MRI were included in the RESPONSE substudy, resulting in a total of 500 patients. Extensive biomarker analysis (sTILs, ROR scores and molecular subtyping by PAM50, and cancer metabolism gene expression) was specifically performed in a matched cohort of 21 PET [−] and 21 PET [+] criteria patients. Figure 1 shows the patient disposition.

Fig. 1
figure 1

Patient disposition

Full size image

Among patients in this substudy, the median age was 52 years (range, 20–83), 42.2% (211/500) had node-positive disease, 68.2% (341/500) had HR-positive tumors, and 75.2% (376/500) had HER2 3+ tumors by immunohistochemistry. Median SUVmax at baseline was 7.2 (range, 1–39.3), and median tumor size by MRI was 33 mm (range, 15.3–157) (Table 1).

Table 1 Clinicopathological characteristics (analysis of the entire population and by [18F]FDG-PET status)
Full size table

Correlation between SUVmax at baseline and clinicopathological characteristics in all patients

SUVmax at baseline was higher in tumors with stage IIIA (p < 0.01), a diameter ≥ 2 cm (p < 0.01), lymph node involvement (p < 0.01), HR-negative status (p = 0.032), higher HER2 protein expression (p < 0.01), increased Ki67 proliferation index (p = 0.01), higher histological grade (p < 0.01), and ductal carcinoma type (p = 0.013) (Fig. 2).

Fig. 2
figure 2

Association between SUVmax at baseline and clinicopathological characteristics

Full size image

Association between [18F]FDG-PET status and clinicopathological characteristics in all patients

Median SUVmax at baseline was 2.7 (range, 1.0–4.41) and 8.0 (range, 2.1–39.3) in PET[−] criteria (screening failures due to the lack of ≥ 1 breast lesion evaluable by [18F]FDG-PET) and PET [+] criteria patients (included) (p < 0.01), respectively. Median tumor size by MRI was 32 mm (range, 16–100) and 33 mm (range, 15.3–157) in PET [−] and PET [+] criteria patients (p = 0.231), respectively (Table 1).

In an unadjusted analysis, PET [−] criteria patients showed more early-stage tumors (p < 0.001), decreased tumor size (p = 0.014), absence of lymph node involvement (p < 0.01), more non-ductal histology (p = 0.013), and lower histological grade (p < 0.01) (Table 1).

Using an adjusted analysis, we selected patient’s characteristics more likely to predict [18F]FDG-PET status to match PET [−] and PET[+] criteria patients. PET [−] criteria patients had a lower tumor stage (odds ratio (OR) 3.0, 95% CI 1.1–4.1; p = 0.014), smaller tumor size (OR 2.2, 95% CI 1.1–4.1; p = 0.014), absence of nodal involvement (OR 4.2, 95% CI 2.2–8.7; p < 0.01), and lower histological grade (OR 0.32, 95% CI 0.16–0.6; p < 0.01) (Table 1).

sTILs, PAM50 intrinsic subtyping and ROR scores, and cancer metabolism gene expression according to [18F]FDG-PET status in matched cohorts

After matching for tumor size, lymph node involvement, and histological grade, differences in sTILs and gene expression by PAM50 (intrinsic subtyping and ROR scores) and Vantage 3D™ Cancer Metabolism Panel were analyzed in a cohort of 21 PET[−] and 21 PET[+] criteria patients (Table 2 and Fig. 1).

Table 2 Clinicopathological and biological characteristics for matched samples by [18F]FDG-PET status
Full size table

No differences in the level of sTILs was found among matched PET [−] and [+] criteria patients with a median score of 10% and 15% in PET [−] and PET[+] criteria patients (p = 0.73), respectively (Fig. 3). The percentages of patients in the different levels of sTILs were the same regardless of the [18F]FDG-PET status ([−] or [+]): 66.7% (14/21) of the patients had low sTILs (sTILs < 30%), 28.6% (6/21) intermediate sTILs (sTILs 30–75%), and 4.8% (1/21) high sTILs (sTILs ≥ 75%).

Fig. 3
figure 3

Stromal tumor-infiltrating lymphocytes, HER2-enriched subtype, and risk of recurrence scores according to [18F]FDG-PET status (− / +)

Full size image

PET [−] criteria patients had slightly lower ROR-S scores than PET[+] criteria patients (median, 63.9 (range 11.3–75.7) vs. 64.2 (range, 39.7–78)) and a lower proportion of HER2-enriched subtype (66.7% (14/21 PET [−] criteria) vs. 81.0% (17/21 PET[+] criteria) by PAM50 (p < 0.05). No significant differences for ROR-P scores were observed according to [18F]FDG-PET status (p = 0.63) (Fig. 3).

Genes involved in glucose metabolism (DLAT, IDH2, LDHA, PGK1, PGLS, and TPI1), hypoxia signaling (HIF1A), and carbon metabolism (SLC7A5 and SLC16A3) were under-expressed in PET [−] criteria patients, whereas genes involved in the mTOR pathway (AKT2) and growth factor receptor (FLT3) were overexpressed compared to PET[+] patients (false discovery rate q < 0.05) (Fig. 4).

Fig. 4
figure 4

Box plot of cancer metabolism gene expression according to [18F]FDG-PET status (− / +)

Full size image

Discussion

The PHERGain trial [5] only included patients whose breast tumors were ≥ 1.5 cm in diameter (MRI or ultrasound) in order to reduce screening failures due to the absence of evaluable breast lesions by [18F]FDG-PET. However, despite this inclusion criterion and the aggressive behavior of HER2+ tumors, around 15% of the patients were excluded as a result of the lack of  ≥ 1 breast lesions evaluable by [18F]-FDG-PET [26].

[18F]FDG-PET response after two cycles of treatment was critical for treatment decision-making in patients included in the group of the PHERGain trial not receiving chemotherapy [5]. The adaptive design of this trial, therefore, made it necessary to select patients with breast lesions identified by [18F]FDG-PET, defined as SUVmax ≥ 1.5 × SUVmean liver + 2 SD. This strict requirement is the main factor responsible for this significant rate of screening failures. For this reason, a better understanding of clinical and molecular determinants of [18F]FDG-PET disease detection could help to better select patients for studies that use [18F]FDG-PET as assessment method.

The median SUVmax at baseline in our entire analyzed population was 7.2 (range 1–39.3), which is similar to previous studies including patients with HER2+ tumors [19, 27]. Additionally, our findings are in line with previous studies in patients with EBC that have demonstrated a relationship between SUVmax and several classical clinicopathological characteristics such as clinical stage [14], HR status [15, 16], Ki67 proliferation index [11, 17], tumor size [18], and histological grade [19,20,21] regardless of BC subtype. Nevertheless, our results are of particular interest because they were specifically generated in patients with HER2+ tumors.

One of the main objectives of this substudy was the analysis of the differences in sTILs and gene expression by PAM50 (intrinsic subtyping and ROR scores) and Vantage 3D™ Cancer Metabolism Panel in a matched cohort of excluded and enrolled patients in the PHERGain trial based on the [18F]FDG-PET inclusion criteria. TILs are predictive biomarkers of response to neoadjuvant therapy in patients with HER2+ tumors [28, 29]. Interestingly, we did not find differences in the levels of sTILs among matched PET [−]/[ +] criteria patients.

Compared with PET[+] criteria patients, PET [−] criteria patients had lower ROR scores, a prognostic factor that has been considered superior to other classical clinicopathological characteristics [30]. Lower ROR scores are associated with a reduced risk of BC relapse in patients with HR+/HER2- EBC [30]. However, the prognostic role of ROR scores in HER2+ tumors remains undetermined.

Our results also showed a higher proportion of PAM50 HER2-enriched tumors and higher levels of HER2-protein expression by immunohistochemistry in PET[+] criteria patients. HER2-enriched subtype and HER2 3+ tumors by immunohistochemistry appeared to be associated with higher pCR rates following anti-HER2-based regimes. These findings are consistent with the capacity of [18F]FDG-PET to predict a pCR to neoadjuvant treatment with HER2-targeted therapies [31].

Regarding the gene under-expression we observed in PET[−] criteria patients, with low [18F]FDG avidity, the lower expression of HIF1A was in concordance with the higher SUVmax detected in patients with BC with high HIF-1A expression [32]. On the other hand, the reduction of glucose metabolism, suggested by the lower expression of genes involved in glucose metabolism, could be justified with the fact that tumor cells can switch their metabolic pathway from glucose to other nutrients such as glutamine [33, 34].

Our substudy has five main limitations worth noting. First, its exploratory nature; the results reported here should be interpreted with caution and considered hypotheses-generating. Second, there was a small sample size in the matched cohorts of PET [−] and PET[+] criteria patients (21 patients each); analyses with a larger sample size would have provided more consistent results. Third, we have tested many variables in a small population and the use of a false discovery rate only partially addresses this limitation. Fourth, matching could decrease external validity because the controls become more similar to the cases than expected in the target population. Fifth, due to the high partial volume effect of [18F]FDG-PET, the SUV of tumor lesions < 2 cm may be artificially reduced.

Conclusions

Our findings highlight the clinical, biological, and metabolic heterogeneity of HER2+ breast cancer, which may facilitate to select HER2+ EBC patients likely to benefit from [18F]FDG-PET imaging as a tool to guide therapy.