Pretreatment ¹⁸F‐FDG uptake heterogeneity may predict treatment outcome of combined Trastuzumab and Pertuzumab therapy in patients with metastatic HER2 positive breast cancer

Breast cancer(BC) is the most prevalent malignancy, accounting for 31% of female cancer cases, and the second-leading cause of cancer death in women [1]. Human epidermal growth factor receptor 2(HER2) is a receptor tyrosine-protein kinase, which is highly expressed in about 15% of patients with BC [2, 3]. This subtype of BC is considered aggressive and is associated with a poor prognosis [4]. However, monoclonal antibodies against HER2 protein can significantly improve the prognosis of patients with HER2 positive BC [5]. Research shows that Trastuzumab mainly binds to subdomain IV of HER2 extracellular domain, while Pertuzumab mainly binds to subdomain II of HER2 extracellular domain to exert their antitumor effect [6, 7]. The combination therapy with Trastuzumab and Pertuzumab(TP) in the treatment of MHBC patients can combine different HER2 epitopes to more comprehensively target HER2 receptor, which produces greater antitumor activity than either agent alone [8]. A recent double-blind, randomized, placebo-controlled phase 3 clinical trials showed that an 8-year overall survival rate was 37% for patients with MHBC treated by the dual-target therapy with TP [9].

At present, the dual anti-HER2 therapy with TP and docetaxel/paclitaxel is the first line of treatment for HER2 positive metastatic or recurrent breast cancer according to the National Comprehensive Cancer Network (NCCN) Guidelines Version 4.2023 [10]. However, there are still some patients with MHBC who cannot benefit from TP. There is an urgent need for biomarker to early select patients with HMBC who can benefit from anti-HER2 therapy before treatment for oncologists.

As a functional imaging technique, PET/CT has been widely used in the field of clinical oncology, with indications including tumor diagnosis, treatment response evaluation, recurrence monitoring, etc. [11‐13]. Some studies have shown that pretreatment ¹⁸F-FDG PET/CT has a good predictive value in evaluating the efficacy of treatment for tumors, and the predictive value of intra-tumoral heterogeneity is particularly prominent [14]. PET/CT parameter heterogeneity index(HI) was first proposed by Johannes Salamon, which was applied to evaluate the characterization of peripheral nerve sheath tumors in neurofibromatosis [15]. Research shows HI has been shown to have good prognostic efficacy in various tumors in many studies [16‐19]. We have previously showed that the HI of ¹⁸F-PET/CT uptake can potentially predict the treatment outcome, including breast cancer [16, 20], nasopharyngeal carcinoma [14, 17, 18]. However, the HI of ¹⁸F-FDG PET/CT imaging has not been reported to predict the prognosis of patients with MHBC treated by the TP, and its predictive value is unclear.

Hence, the aim of this study is to evaluate the predictive ability of heterogeneity in ¹⁸F-FDG uptake to predict the therapeutic outcomes of dual target therapy with TP for MHBC.

In this retrospective study based on baseline pretreatment ¹⁸F-FDG PET/CT imaging to early allow prediction of treatment outcomes with TP for MHBC, we determined that the intra-tumoral HI based on PET/CT parameters was a predictive factor for PFS in patients MHBC treated by TP, and low HI at baseline was significantly associated with improved PFS. As far as we know, this is the first study to evaluate the predictive ability and prognostic significance of ¹⁸F-FDG PET/CT in patients with MHBC treated by TP, which provides potential clinical value for clinical individualized treatment.

In order to accurately predict the prognostic factors of dual target therapy for MHBC, Kazuhiro Araki et al. showed that baseline absolute lymphocyte counts were a predictive factor for PFS in MHBC treated by TP [22], but infection and prior chemotherapies both may affect baseline levels of absolute lymphocyte counts, therefore, the absolute lymph nodes in predicting PFS of patients with MHBC is controversial. Meanwhile, Koji Takada et al. also showed that tumor-infiltrating CD8 to FOXP3 lymphocyte ratio may predict treatment responses of dual target combination therapy [5], but which is invasive and difficult to be accepted by patients. Connolly et al. showed that the early changes in tumor maximum standardized uptake value corrected for lean body mass (SULmax) of PET/CT parameter predict pathologic complete response to neoadjuvant TP in stage II or III of patients with estrogen receptor-negative and HER2-positive breast cancer [23]. Therefore, we tried to investigate pretreatment ¹⁸F‐FDG uptake heterogeneity predicts the outcome of patients with MHBC treated by TP.

¹⁸F-FDG PET/CT imaging, as the most common examination method for tumor patients, is also applied to the diagnosis, staging, monitoring and efficacy prediction of patients with breast cancer [20, 24, 25]. The most common PET/CT parameters, including SUVmax, MTV and TLG were used in various clinical scenarios [26‐28]. Unfortunately, there is no correlation between above common parameters and PFS of patients with MBC treated by TP in this study. As mentioned above, HI has been shown to have good prognostic efficacy in various tumors in many studies [16‐19]. In the early stage, we tried to predict the treatment results of first-line chemotherapy in patients with metastatic triple negative breast cancer with HI from pretreatment ¹⁸F-FDG, and obtained satisfactory result [20]. And ¹⁸F-FDG uptake heterogeneity was applied to reflect spatial tumor heterogeneity among metastases, and baseline HI-intra and HI-inter could both predict the treatment efficacy of pyrotinib in patients with HER2 positive BC [29].

Gradient PET/CT parameters are often used for clinical correlation and predictive analysis [21, 30]. In this study, we analyzed the relationship between gradient PET/CT parameters and the progress of HMBC. And the results show that only the HI_40% and HI_50% had significant negative correlations with the progression of HMBC after double target therapy. The main reasons may be as follows: as the cut-off value for distinguishing benign and malignant pulmonary nodules in SUVmax, 2.5 was proposed by a small sample size study in 1993 [31], which cannot be widely used for other tumors, including MHBC. 60% and 70% SUVmax fixed threshold segmentation may lose more tumor cell metabolic data on the one hand, and may reduce volume of interesting (VOI) and increase partial volume effect on the other hand.

Some studies show that there are differences in individual treatment responses, which has been attributed to intra-tumoral heterogeneity, including tumor cell metabolism, proliferation, angiogenesis, necrosis, and tumor fibrosis [14]. Other studies show that immune cells in primary tumors is very important for tumor treatment, which can predict the tumor treatment efficacy [32]. Therefore, individual variability in tumor therapeutic response is not only due to the heterogeneity of tumors, but also may be affected by the immune cell infiltration of tumors. In the past, it is generally believed that ¹⁸F-FDG is taken up by tumor cells in tumor lesions, and the heterogeneity of ¹⁸F-FDG uptake in tumors was also affected by the number of tumor cells, proliferation, angiogenesis, necrosis and hypoxia. However, the latest research shows that¹⁸F-FDG is not only taken up by tumor cells, but also by immune infiltrating cells in tumor lesions [33]. Therefore, ¹⁸F-FDG uptake heterogeneity can more comprehensively show the heterogeneity of tumor cells and immune cells in tumor lesions. This may be one of the reasons why HI can predict PFS early of patient with MHBC in this study, but they still need to be confirmed by corresponding mechanism experiments.

With the rise of artificial intelligence, more and more machine learning algorithms, including Artificial neural network (ANN) [34], Decision tree (DT) [35], K-nearest neighbour (KNN) [36], Logistic regression (LR) [37], Naïve Bayes (NB) [36], Random forest (RF) [38] and Support vector machine (SVM) [39], are used to predict clinical treatment [40, 41]. Although the above machine learning algorithms have their own advantages in the prediction process, they are all too complex and require the assistance of specialized computer engineers, which is not practical in clinical practice. However, the HI mentioned in this study represents the metabolic distribution within the tumor, which is very easy to obtain by dividing SUVmax by SUVmean. At the same time, it is easy for clinicians to understand and accept. Therefore, the clinical practicability of the HI is significantly higher than the above computer learning algorithm. Taking this study as an example, the higher the PET/CT parameter-HI, the worse the dual-target therapy effect be in patients with MHBC, which implies whether clinicians need to increase other supplementary treatments or to improve the effectiveness of therapy or directly switch to second-line treatment.

There are still some shortcomings in this study. First, some HER2 imaging agents have been studied in clinic, including ⁸⁹Zr-trastuzumab [42] and ⁶⁸ Ga-HER2 affinity [43],¹⁸F-nanobody [44] etc. However, which are not mature and still in the stage to evaluate whether tumor uptake of HER2 imaging can distinguish HER2-positive from HER2-negative breast cancer. Second, this study is a small sample regression exploration and the number of events is not large. Finally, as the data of HI proposed in this study are mainly distributed between 1 and 2, so that cox proportional hazards analysis was not conducted in this study.

Our preliminary study showed that HI based on pretreatment ¹⁸F-FDG uptake may predict the PFS of patients with HER2 positive MBC treated by first line dual target therapy. It could be useful to identify patients with HER2 positive MBC who would benefit from dual target therapy and provide theoretical support for individualized treatment. Although this preliminary study outlines the potential prognostic performance of HI, findings need to be further evaluated and confirmed by large, prospective clinical research.