Introduction
Lung cancer has the highest morbidity and mortality of all malignant tumours, which is one of the most common cancers worldwide [
1]. The major histologic subtypes of lung cancer are non-small cell lung cancer (NSCLC, 85%) and small cell lung cancer (SCLC, 15%) [
2]. SCLC is an invasive, high-grade malignancy with early development and a bad prognosis. Most patients with SCLC are advanced with widespread metastasis when they are diagnosed, and systemic chemotherapy is the most effective therapy [
3]. Few therapeutic methods are available after SCLC relapse, and the prognosis is especially poor. Therapeutic treatments for patients with SCLC have essentially remained unchanged in recent years [
4]. Thus, further research into the mechanism of SCLC and the exploration of new therapeutic targets for SCLC is imperative.
Delta-like protein 3 (DLL3) is a transmembrane protein that promotes the development of neuroendocrine tumours through its reciprocity with the Notch pathway [
5]. It is expressed in cancer tissues in approximately 85% of patients with SCLC and other neuroendocrine cancer but is not expressed in non-neuroendocrine tumours or normal tissues [
6]. Most studies have shown that DLL3 is a latent treatment target point for SCLC [
7‐
9]. Recently, many researches initiate to focus on the prognostic value of DLL3 in SCLC. Some of the researchers demonstrated that DLL3 had no correlation with prognosis in SCLC patients [
10,
11], but other investigators showed that DLL3 was associated with the survival of patients with SCLC [
12]. So previous studies of the prognostic value of DLL3 in SCLC remain controversial, we performed this systematic meta-analysis for resolving this controversial.
Published articles discussing the prognostic value of DLL3 in SCLC were systematically reviewed in our meta-analysis. We aimed to include all correlational studies to assess the prognostic value of DLL3 and attempted to identify an accurate biomarker to guide prognosis and treatment for SCLC in the future.
Discussion
Many studies have shown that DLL3 played a significant prognosis value in patients with cancer. For example, high DLL3 expression was associated with shorter OS and PFS in small cell bladder cancer [
26]. High DLL3 expression was associated with bad OS and usually expressed in older patients and advanced stage in endometrial cancer [
27]. Xie et al. reported that high DLL3 expression predicted a better OS in patients with SCLC [
21], and other studies foreboded that high DLL3 expression was an inferior prognostic marker for SCLC [
20,
23]. Thus, the results of different studies remained controversial. We included a total of 5 studies with 601 patients with SCLC to assess the prognostic value of DLL3 in SCLC by pooling the data of HR and 95% CI.
First, we performed this meta-analysis using 6 studies. But heterogeneity was observed in our pooled results. Through further subgroup analysis, we found that a study from the American was the main cause of heterogeneity. As we removed the American study, we found that high expression of DLL3 is a marker of poor prognosis in SCLC. But we also noticed that the results of the American study were contrary to our conclusions, they reported that high expression of DLL3 is a marker of good prognosis in SCLC. One interpretation of this result was that DLL3 expression varies between different populations. The American study indicated that the high expression of DLL3 was a marker of good prognosis [
21], and there was no significant correlation between DLL3 expression and prognosis in several Japan studies [
22,
25], while the high expression of DLL3 is a marker of poor prognosis in China studies [
20,
23]. These results showed that the expression of DLL3 might be different in different populations. However, only one study researched the correlation with DLL3 expression and prognosis in SCLC outside of Asia, so the results need to be treated with caution. In addition, more studies were suggested for the future to further verify the existence of such differences.
Our other explanation of this result was that the expression of DLL3 was the same in different populations. Among the included studies, the sample size of 5 studies was less than 100, including the America and Japan studies. Therefore, we speculated that insufficient sample size might cause bias in the results. We conducted a subgroup analysis according to the sample size, and the results showed that the high expression of DLL3 in the studies with large sample sizes (
N ≥ 100) was associated with poor prognosis, while the pooled results of the studies with low sample sizes (
N < 100) showed no significant correlation between the expression of DLL3 and prognosis. Therefore, the sample size may be one of the reasons for the differences in DLL3 expression in each study. Moreover, immunohistochemistry was used to detect DLL3 expression in all of the studies. Immunohistochemical staining is a semi-quantitative method and is evaluated with great subjectivity [
28]. Different antibodies and different cut-off values for DLL3 expression were employed in all the included studies and thus could also be another cause of the differences in results. The American study also explained their different results by claiming that many of the other studies were performed using mRNA expression instead of protein expression or used different cut-offs value of DLL3 expression [
21].
Researches have reported that DLL3 is highly expressed in SCLC [
29,
30], which suggested that DLL3 might promote the development of SCLC. Therefore, we also discussed the correlation between DLL3 expression and the clinical characteristics of patients with SCLC. The pooled results showed that DLL3 expression had no significant correlation with patients’ sex, smoking status and stage, while DLL3 often highly expressed in metastasis patients of SCLC. Our survival analysis outcomes were consistent with this result, which suggested that high expression of DLL3 might be one of the factors contributing to poor prognosis in patients with advanced metastatic SCLC in Asia. However, only a few studies reported the correlation between the expression of DLL3 and clinical characteristics, and the methods and cut-offs values used to detect the DLL3 expression in each study were not uniform. Therefore, more reliable studies are needed to further verify our outcomes. Our results suggest that it is valuable to further investigate the correlation between DLL3 and the clinical characteristics of patients with SCLC.
Our meta-analysis is the first to focus on the prognostic value of DLL3 in SCLC. The significance of this meta-analysis lies in providing a basic direction and evidence for further research into the mechanism of DLL3 in SCLC. For SCLC, Notch1 over expression could induce G1 cell cycle arrest [
31]. Previous studies reported that DLL3 downregulated the Notch receptor expression, thereby the Notch signalling pathway was inhibited within the cell [
32]. Therefore, high expression of DLL3 can promote the development of SCLC by inhibiting the Notch signalling pathway. Studies also have shown that the high expression of DLL3 may reduce the sensitivity of chemotherapy drugs [
24]. These studies have demonstrated that DLL3 may be associated with the prognosis of SCLC and also consistent with our meta-analysis results. Thus, studies of the corresponding targeted drugs of DLL3 can effectively inhibit the expression of DLL3 and thus improve the survival of SCLC. Rovalpituzumab tesirine (Rova-T) is a new antibody-drug conjugate directed against DLL3 in SCLC [
33]. A phase I trial found that patients with high DLL3 expression in SCLC showed a better response to Rova-T than those with a low DLL3 expression [
34]. However, disappointingly, the phase III TAHOE trial has been stopped because the Rova-T group showed a worse OS compared to the control group [
35]. But more clinical trials are recruiting participator to investigate Rova-T as maintenance therapy in advanced stage SCLC. The lack of progress with this drug does not prevent us from making a breakthrough with other similar drugs. Some researches found that the intra-tumoural and inter-tumoural distribution of DLL3 protein in SCLC is homogeneous [
20], supporting the conclusion that biopsy specimens are a reliable source for DLL3 evaluation for targeted therapy. In addition, most studies have demonstrated that DLL3 is highly expressed in SCLC, while it is not or is less expressed in other types of lung cancer and normal tissues [
36]. Therefore, the expression of DLL3 can be detected by biopsy as an indicator for diagnosis, predicting therapeutic efficacy and monitoring recurrence or metastasis of SCLC in the future.
Although our study fully explains the prognostic value of DLL3 in SCLC, our analysis still has several limitations. First, large heterogeneity was observed in the pooled results. This is explained by the observation that the evaluation criteria for the expression of DLL3 are particularly mixed, and there are no international standards for cut-offs values to determine the expression of DLL3. Thus, the scoring methods and cut-offs values of DLL3 should be unified to strengthen our conclusions. Otherwise, the detection method of DLL3 in most studies is mainly immunohistochemistry at present, which is a semi-quantitative, subjective and inaccurate detection method. Different studies show different prognostic values of DLL3. Therefore, we need other more precise detection methods to evaluate the expression of DLL3 in SCLC in the future.
Second, the therapy method is also a key limitation. The current studies only focus on tissue specimens from patients with SCLC after surgery or biopsy, and few studies reported the treatment methods in their researches. However, the prognostic value of DLL3 may lie in the therapeutic method. Therefore, every study should pay attention to the impact of patient treatment methods on prognosis in the future.
Third, some of the original studies did not report the data of HR and 95% CI. The HR and 95% CI results were measured from survival curves, an evaluation method with certain deviation and subjectivity, which might influence the authenticity of the results.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.