The incidence of MPE is high in patients with breast cancer, with approximately 7–11% of breast cancer patients suffering from MPE during the course of the disease.[12] The mechanisms for the development and progression of MPE are highly complex and may be related to a range of factors, including impaired lymphatic drainage and pleural invasion.[13] Previous studies have illustrated that the humoral and mechanical milieus of the body are substantially altered after the development of MPE. Patients with MPE often report symptoms of cough, chest tightness, and breathlessness if their presentation is compounded by dysfunction of the heart, lungs, and other organs, which can easily result in a poor prognosis. Currently, guideline recommendations on therapy after the diagnosis of MPE are not uniform and focus mainly on the relief of symptoms.[3, 14–16] A possible reason is that such patients have highly variable outcomes, which prevents physicians from reaching a more accurate assessment of the prognosis.[17, 18] Several prognostic models aiming to predict the OS of pancancer patients with MPE have emerged thus far. However, the numbers of breast cancer patients included in the analyses were relatively small.[5–7] Since then, several models focusing on lung cancer and mesothelioma have been developed.[19, 20] With breast cancer as the second leading cause of MPE, it is essential to construct a model for prognostication in a cohort of breast cancer patients. Based on the clinical characteristics and laboratory indicators determined via the first thoracentesis, we constructed a nomogram to predict the survival probability of individual breast cancer patients with MPE for the first time. The modeling data sources are derived from real-world healthcare data, and a wide range of variables can be collected.
Our results reconfirmed that MPE confers a poor prognosis in breast cancer patients, with substantial heterogeneity. The model was built using univariate Cox regression and LASSO regression analyses. The latter is considered to be able to screen multiple variables in cases of relatively small sample sizes and has been widely used in the construction of prognostic models.[21, 22] In our final nomogram, eight prognostic variables were retained. Several of these have previously been shown to be significantly relevant to the survival of cancer patients, including the EP/PR status, NLR, and DFS.[23–25] Whether the levels of biochemical indicators in the PE can serve as a predictor in MPE prognosis has been controversial and the cutoff values of these markers have varied widely among studies.[26–29] In the present study, laboratory indicators were included as continuous variables in the analyses, which increased the statistical power. As shown in the mentioned nomogram, glucose in PE contributed the most to the predictive capability of the model, with patients with higher PE glucose levels showing longer survival. However, previous in vitro experiments have shown that a high-glucose environment can promote tumor growth, and in contrast with our findings, a study of MPE in lung cancer reported that lower PE glucose levels (60 mg/dL) predicted shorter survival.[30] The results are still inconclusive. Notably, the presence of ascites and non-first recurrence acted as risk factors in our model. Few studies concerning these two factors have been performed, and it is not clear if this result is specific to MPE in breast cancer. Otherwise, the LENT score, based on the LDH level in pleural fluid, is considered to be associated with the outcome.[5] However, this association was not confirmed by the results of our study and deserves further investigation.
Breast cancer represents a heterogeneous group of tumors that were originally classified into distinct molecular subtypes by their clinicopathological features. It has previously been shown that there are important differences in the metastatic behavior of breast cancer subtypes. However, there are similar odds of metastasis to the pleural and peritoneal regions.[31] In our study, patients with luminal-type tumors had a median survival of 16.87 months, which was significantly better than the median survival observed in patients with nonluminal subtype tumors (7.7 months). In addition, the receptor status included in the study was determined based on the pathological report at the time of initial diagnosis with breast cancer. Some scholars have identified the need to reassess the biomolecular status at metastatic sites because the receptor status is not always in concordance between primary and metastatic breast cancer lesions.[32] Indeed, the rates of discordance in HER2 expression are as high as 63%.[33]
For MPE therapy, an official ATS/STS/STR Clinical Practice Guideline recommends therapeutic pleural interventions only for symptomatic patients with two main types of treatment, i.e., indwelling pleural catheterization (IPC) and pleural fixation.[4, 15] No guideline specifies which is the most appropriate for each individual. However, for a shorter length of hospital stay and fewer pleural procedures, clinicians are more willing to choose IPC.[34] At the time of admission, all patients in our study displayed respiratory symptoms, including dyspnea, cough, and chest pain. After patients were treated with thoracentesis and drainage of 700–1000 ml per day, their clinical symptoms were alleviated to varying degrees. Nevertheless, it needs to be noted that both IPC and pleural fixation are palliative treatments aimed at relieving symptoms. Systemic therapy is still recommended for patients with expected long-term survival. Our study revealed a significant survival benefit for MPE patients treated with systemic or intrapleural chemotherapy in the high-risk group. Therefore, we recommend more aggressive chemotherapy for patients in the high-risk group and cautious use of chemotherapy for patients in the low-risk group.
Our study has several notable strengths. First, our model is specific to the metastatic breast cancer population. Second, the variables in the model are derived from routine laboratory indicators at general hospitals. Third, after the external data were verified, the model proved to predict survival well. Nevertheless, there are also several limitations. First, the sample size was relatively low for a prediction model study, resulting in a certain degree of overfitting. Second, our study was a retrospective study. Patient retrospective bias in the recall of physical activity level can lead to inaccuracy of the ECOG score. Therefore, the ECOG score, which has been shown to be significantly relevant to the survival of MPE patients, was not included in our analysis but has been included in other models of MPE.[9, 23] The absence of this information in our model limits comparison with other models. Third, while we have entered a new era of molecularly targeted therapy, currently, MPE still lacks efficient targeted drugs, and the current study is still at the clinical level. Our prediction model does not include biomarkers that may help to improve the performance of the nomogram model, and more biomarkers need to be further investigated.
In conclusion, our model enables the prediction of 3-, 6-, and 12-month survival probabilities in breast cancer patients with MPE in a manner that is clinically convenient and reliable. We also provide a web calculator to calculate the risk score (https://mpenomogram.shinyapps.io/MPE_Nomogram/). However, more cohort or prospective studies will still be required to confirm our results and credibility of the model in the future.