Chromogranin A is a predictor of prognosis in patients with prostate cancer: a systematic review and meta-analysis

Background

Prostate cancer (PCa) is the second leading cause of cancer death in men with a worldwide annual mortality of over 200,000.^1–3 The first symptom occurs mostly during the progression of PCa when it cannot be treatable anymore. Therefore, early detection of PCa in asymptomatic men is the key to reducing mortality. Prostate-specific antigen (PSA)-based screening detects substantial clinically insignificant patients with PCa.^4,5 However, most of these asymptomatic patients are treated unnecessarily and subject to side effects of treatment. Therefore, population-based PSA screening is not recommended by current guidelines.⁶ In order to distinguish between indolent and high-risk PCa, reliable biomarkers are needed for the better prognostication of PCa.

Chromogranin-A (CHGA) is a common glycoprotein expressed in neuroendocrine cells. Neuroendocrine activity can also be detected in other tumors except for neuroendocrine tumors, such as prostate and breast tumors.^7–10 One study demonstrated that CHGA levels are significantly elevated in patients with castration-resistant prostate cancer (CRPC) in immunohistochemical studies.¹¹ As tumors develop in neuroendocrine tissues, they become the main source of circulating CHGA which can be detected in blood and tissues.^12,13 Moreover, elevated circulating CHGA levels have been confirmed to be a helpful biomarker for the diagnosis of different types of neuroendocrine tumors.^14–17 Some researches have shown that high serum CHGA levels are associated with advanced stage disease and poor prognosis,^18,19 but its practicability and prognostic value in clinically localized PCa is still debatable.

Several studies have suggested that the high mRNA expression of CHGA is related to worse survival, but other studies did not reach these conclusions.^20–23 We thus performed a meta-analysis to reveal the prognostic significance of CHGA expression for overall survival (OS) and progression-free survival (PFS) in patients with PCa.

Materials and methods

This study was performed in accordance with the Cochrane Collaboration criterion.²⁴ For reporting, we followed the PRISMA statement guidelines.²⁵ Thus, no ethical approval and patient consent are required.

Literature search

To identify eligible studies, we performed a comprehensive literature search in the electronic databases of MEDLINE, Embase, Web of Science, and Cochrane Library databases for eligible studies regarding the association of CHGA expression with survival outcomes in PCa from database inception up to September 2018. Each database was searched without restrictions to languages, publication types, or regions using the following combination of Medical Subject Headings (MeSH) and non-MeSH search terms: (“chromogranin A” OR “parathyroid secretory protein” OR “CHGA”) AND (“prostate cancer” OR “prostatic cancer” OR “prostate neoplasm” OR “prostatic neoplasm”). The main search was completed by the senior author (ZLG). Any discrepancy was resolved through consultation of an investigator (SSW) not involved in the initial procedure (as shown in Supplementary Material).

Inclusion and exclusion criteria

Two independent investigators (ZG and YW) selected eligible studies regarding the association of CHGA expression with survival outcomes in PCa without publication status or language restrictions in accordance with the following inclusion criteria: 1) diagnoses were confirmed as PCa by pathology; 2) original trials regarding the association of CHGA expression with survival outcomes (eg, OS, PFS, etc.) in PCa; and 3) studies reporting sufficient data of risk estimates with corresponding 95% CIs or enough data to calculate them directly were eligible for inclusion. If more than one study was identified from the duplicated database, we retained only the most recent or largest study to avoid duplication of information. The exclusion criteria were as follows: 1) reviews, letters, conference papers, case reports, case series, and expert opinion articles were not eligible for inclusion; 2) absence of detailed results; and 3) animal studies. Any disagreement was resolved through adjudication of senior authors.

Data extraction and methodological quality assessment

Two reviewers (ZG and SX) extracted the data independently using a predefined data extraction form, and discrepancies were settled through a consensus discussion. The following data were extracted into a standardized evidence table: first author, year of publication, country, study design, baseline population characteristics (ie, mean age and sample size), duration, adjusted factors, and HRs with corresponding 95% CIs for survival in each comparisons. We also checked these data for accuracy. Moreover, we contacted, if possible, the primary authors of studies with insufficient information to acquire and verify the data.

The quality and risk of bias of the included cohort studies were evaluated by two independent reviewers (FLC and SW) separately according to the Newcastle-Ottawa Scale (NOS).2⁶ Disagreements were also settled through discussion among authors. The standards consist of 10 items that assess the representativeness of the included studies. Each item was evaluated as “yes,” “no,” or “unclear” corresponding, respectively, to “1,” “0,” or “0” according to the information provided by the studies. The total score ranged from 0 to 10 with the overall score categorized as follows: 7 to 10: “high quality,” 4 to 6: “moderate quality,” and 0 to 3: “low quality.”

Statistical analyses

For meta-analysis, the total risk estimate of extracted data was pooled using HRs with associated 95% CIs through the STATA statistical package (version 14.0; serial number: 10699393; StataCorp Wyb) to determine the association of CHGA expression with survival outcomes in PCa. For consistent definitions, HRs with corresponding 95% CIs were used as a common measure in the included studies because CHGA expression in PCa was considered as a rare event. We could generally ignore the distinctions among the various measures of risk estimates. Thus, the ORs values in the observational studies were considered as approximations of HRs.²⁷ The aggregated results and 95% CIs for effect size were calculated using inverse-variance weighted meta-analysis. The I-square (I²) test was performed to assess the effect of study heterogeneity on the meta-analysis results, with I² values of 0%, 25%, 50%, and 75% representing no, low, moderate, and high heterogeneity, respectively. Based on the Cochrane review guidelines,²⁴ a severe heterogeneity of I²≥50% warrants the use of random-effects models. Otherwise, a fixed-effects model is utilized. Statistical significance was set at P<0.05. Subgroup analyses were performed in accordance with different countries, study designs, and chromogranin-A level measurements. Sensitivity analysis was conducted by deleting each study individually to evaluate the quality and consistency of the results. A meta-regression analysis was conducted to investigate possible sources of heterogeneity on certain variable. The restricted maximum likelihood method was used for analysis. Finally, Egger’s and Begg’s tests were performed to assess publication bias, and the funnel plot symmetry was examined.^28,29

Results

Study selection process

A flow chart depicts the search process and study selection (as shown in Figure 1). A total of 665 studies were identified through our comprehensive search of several electronic databases. Only 594 studies were retrieved after removal of duplicates. After screen the titles and abstracts, only 28 studies remained. Finally, a total of 18 full-text articles were discarded for the following reasons: survival was not the end point (6 studies), men did not match the PCa definition (3 studies), duplicated database (1 studies), and no sufficient data for extraction (8 studies). Therefore, 10 cohort studies^{20–23,30–35} comprising 3,172 participants were included in our meta-analysis in compliance with the inclusion criteria.

Figure 1 Flow diagram of literature searches according to the preferred reporting items for systematic reviews and meta-analyses statement.

Study characteristics and methodological quality

The basic characteristics of the included studies^{20–23,30–35} are described in Table 1. These studies (7 retrospective cohort studies^{20,21,23,30–33} and 3 prospective cohort studies^22,34,35) were published between 2005 and 2018 and involved 3,172 participants. In total, 10 eligible cohorts with 3,172 patients were analyzed for OS outcomes,^{20–23,30–35} and 3 qualified studies with 139 patients were analyzed for PFS outcomes.^21,31,35 The sample sizes varied from 39 to 1,018. Among the included studies, five were conducted in Italy,^{20–22,30,31} two in Germany,^33,35 two in USA,^32,34 and one in UK.²³ Furthermore, serum CHGA levels were measured in six studies,^{21–23,31,32} tissue CHGA levels were measured in one study²⁰ and plasma CHGA levels were measured in three studies,^30,33,34 respectively. All studies were published in English,^{20–23,30–35} and nine studies^{20,21,23,30–35} provided adjusted HRs and one²² reported ORs for survival, accounting for confounding factors. Moreover, the duration and the mean age of patients were also detailed in Table 1.

Table 1 Characteristics of the included studies

In addition, the methodological quality was evaluated in accordance with NOS.²³ It was found that eight studies^{20,23,31–35} got 8 points, respectively, one study22 got 6 points, and one study21 obtained 5 points. In all, the methodological quality of eight studies was considered of high quality,^{20,23,31–35} and two studies were evaluated as moderate quality.^21,22 The main deficiency was the selection bias related to the insufficient adjustment of clinical tumor stage and initial treatment modality among the included studies.

CHGA expression and OS in PCa

Ten cohort studies involving 3,172 patients provided the data on OS in PCa.^{20–23,30–35} The results showed an association between high CHGA expression and worse OS (HR=1.24, 95% CI: 1.07–1.44; P=0.004) in PCa. Significant heterogeneity was observed among studies (I²=77.6%, P<0.001), and a random effects model was used for pooled analysis (Figure 2). Sensitivity analysis indicated that none of the individual studies influenced the summary statistic by omitting single study from the pooled estimate substantially (Table 2). As summarized in Table 3, we conducted subgroup analyses regarding different study designs, showing that the summary HR was statistically significant in the retrospective cohort studies (HR=1.51, 95% CI: 1.14–1.99; P=0.004; I²=72.4%) rather than prospective cohort studies (HR=1.08, 95% CI: 0.95–1.23; P=0.257; I²=69.1%). Additionally, as for subgroup analyses of the different countries and CHGA levels measurements, the results were inconsistent with overall analysis except for the plasma CHGA level measurement, USA, and Germany on account of the limited number of studies evaluated. Literature reports of potential impact on tissue CHGA levels in PCa are scarce and also with contrary results. Thus future studies regarding tissue CHGA levels in PCa would verify our results. Notably, when stratified by clinical tumor stage, the results showed that the summary HR was statistically significant in patients with CRPC (HR=1.29, 95% CI: 1.07–1.55; P=0.008; I²=57.9%) compared with patients with PCa (HR=1.23, 95% CI: 0.93–1.63; P=0.148; I²=81.7%). However, the subgroups could not find the potential factors that may substantially affect the heterogeneity. Meta-regression analysis results shown that none of the covariates (country, P=0.826; study design, P=0.295; CHGA levels measurements, P=0.415; clinical tumor stage, P=0.762) resulted in heterogeneity among the included studies. Therefore, the adjusted R-squared of −57.20% to −27.62% what expresses is that the regressors are contributing little to the explanation of the response variables (Table 4). Finally, potential publication bias might exist because of the small number of studies evaluated (Egger’s test, P=0.001; Begg’s test, P=0.020) (Figure 4).

Table 2 Results of sensitivity analyses

Table 3 Results of subgroup analyses

Table 4 Results of meta-regression

Figure 4 Funnel plot of the HR of CHGA expression and overall survival in prostate cancer. The Egger regression asymmetry plot shows the standardized effect estimates versus precision, the variance-weighted regression line, and the CI for the intercept. Individual studies are represented by white circles. Failure of this CI to include zero indicates asymmetry in the funnel plot and may give evidence of publication bias. Guidelines at x=0 and y=0 are plotted to assist in visually determining if zero is in the CI. Abbreviation: CHGA, chromogranin-A gene.

Figure 2 CHGA expression and OS in PCa. Individual studies are represented by black squares and horizontal lines that correspond to the point estimate and 95% CI of the OR. The size of the black square reflects the weight of the study in the meta-analysis. The solid vertical line corresponds to “no effect” of treatment – an HR of 1. The diamond at the bottom and the dotted line represent the combined or pooled HR of all 10 trials with their 95% CI. Abbreviations: CHGA, chromogranin-A gene; OS, overall survival; PCa, prostate cancer.

CHGA expression and PFS in PCa

Three cohort studies involving 139 patients provided data on PFS in PCa.^22,34,35 The combined analysis of three studies suggested that high CHGA expression was not significantly associated with shorter PFS (Figure 3), with HR of 1.73 (95% CI: 0.92–3.28; P=0.090; I²=73.9%). However, the use of subgroup analysis and sensitivity analysis were limited because of the small number of studies evaluated.

Figure 3 CHGA expression and PFS in PCa. Individual studies are represented by black squares and horizontal lines, which correspond to the point estimate and 95% CI of the odds ratio. The size of the black square reflects the weight of the study in the meta-analysis. The solid vertical line corresponds to “no effect” of treatment - an HR of 1. The diamond at the bottom and the dotted line represent the combined or pooled HR of all 3 trials with its 95% CI. Abbreviations: CHGA, chromogranin-A gene; PCa, prostate cancer; PFS, progression-free survival.

Discussion

Main findings

In this study, we analyzed the associations between CHGA expression and survival in patients with PCa using a meta-analysis of the 10 included studies.^{20–23,30–35} Overall, our results indicated a significant association of high CHGA expression with poor OS in PCa, although statistically significant difference was not observed in the PFS group. Also, it should be noted that significant heterogeneity was observed in the overall analysis. The meta-regression and subgroup analyses could not find the potential factors that may substantially affect the heterogeneity. The results of sensitivity analysis validated the rationality and reliability of our analysis.

Five studies suggested a significant association between high CHGA expression with poor OS in PCa,^{20,21,30,33,34} whereas five studies reported conflicting results.^{22,23,31,32,35} The retrospective cohort conducted by Burgio et al³¹ reported that in the multivariate analysis, high CHGA expression was a predictor of PFS (HR=2.70, 95% CI: 1.30–5.20; P=0.0047), while CHGA levels remained of borderline significance of OS (HR=1.90, 95% CI: 0.90–4.10; P=0.0919). Moreover, De Nunzio et al²² focused on poorly differentiated PCa, their study indicated that high CHGA expression was not a significant predictor of OS on univariate (OR=1.00, 95% CI: 0.99–1,01; P=0.66). Similarly, Giridhar et al³² Jeetle et al²³ and von Hardenberg et al³⁵ demonstrated that CHGA expression was not an independent predictor of survival (HR=1.35, 95% CI: 0.89–1.97; P=0.154; HR=0.97, 95% CI: 0.89–1.37, P=0.87; HR=1.07, 95% CI: 0.93–1.24, P=0.0919, respectively). When we excluded this study from the meta-analysis, the stability of the results showed no significant changes, validating the rationality and reliability of our analysis.

Implications for clinical practice

CHGA is considered to be a useful predictor in PCa patients with lower PSA levels.^36,37 However, Our results suggest that the lack of correlation between the low PSA levels and high CHGA value, which may be the expression of patients with advanced CRPC with neuroendocrine differentiation in the context of a heterogeneous tumor volume. Moreover, elevated serum CHGA levels are usually associated with neuroendocrine differentiation of CRPC, suggesting that patients with CRPC have a poor prognosis and may be related to the increasing degrees of differentiation.^34,38 In recent years, several studies have suggested that serum levels of CHGA are significantly higher in metastatic PCa as compared to non-metastatic and primary PCas.^39,40 Nevertheless, despite this interest in neuroendocrine serum markers, little is known about their sources in different growth patterns and PCa metastasis. Previous studies indicated that CHGA values were not substantially affected by either endocrine therapy or chemotherapy.^30,41 Since preoperative treatment decisions based on the prognosis may be active surveillance rather than surgery, our results may have direct clinical relevance. In addition, our results show that those who are eligible for active surveillance may also benefit from the analysis of CHGA, as this may improve the accuracy of disease surveillance.

Latest classification of neuroendocrine prostate tumors suggests that variants of neuroendocrine prostate cancer (NEPC) contain a mixed form between conventional adenocarcinoma and NEPC, which was often characterized by androgen receptor independence.⁴² However, neuroendocrine markers and androgen receptor are dually expressed in the mixed tumor cells due to clinical and pathological heterogeneity among patients.⁴² Moreover, a large proportion of the treatment-related recurrent prostate tumors have strong neuroendocrine characteristics, which was caused by pre-existing prostate adenocarcinoma.⁴³ Several studies indicated that neuroendocrine differentiation cells are potentially derived from the common pluripotent stem cell populations, which play a crucial role in CRPC and might contribute to hormone therapy and chemoresistance.^44,45 Notably, NEPC cells express high levels of chromogranin and synaptophysin.⁴⁶ Hence, the overlapping clinical entities require more accurate clinical and molecular classification, and further research is needed to determine their prognostic impact.⁴⁷

Strengths and limitations

In general, our meta-analysis exhibited several crucial strengths. First, the meta-analysis was the first to assess the association between CHGA expression and survival in patients with PCa through thorough systematic search and rigorous analytical approaches. Second, multivariable-adjusted risk estimates were applied to minimize the confounding factors that might influence the whole results. Third, the rationality and reliability of our meta-analysis was observably improved because the overall combined estimates were based on a large sample size. Furthermore, sufficient subgroup analyses and sensitivity analyses were also performed to ensure the reliability of this study.

This study has several limitations, which may affect the interpretation of some of our results. Firstly, there was the problem of heterogeneity in the overall analysis, although we could not find the potential factors that may substantially affect the heterogeneity through the meta-regression and subgroup analyses. Moreover, clinical tumor stage and initial treatment modality should also be accounted for potential considerable factors, which possibly contribute to heterogeneity. However, our understanding of the effects of clinical tumor stage and initial treatment modality among the included studies on the overall results remains insufficient, although these factors have been investigated, albeit inadequately, in other studies. Further research is needed to verify the findings of this meta-analysis with regard to the factors affecting extensive consequences and common objectives as mentioned above. Second, CHGA expression has been associated with other molecular biomarkers for PCa prognosis. Inadequate adjustment for these biomarkers in several included studies might have resulted in spurious associations. Finally, potential publication bias might exist because of the small number of studies evaluated and the small effect size estimated.

Conclusion

In summary, current evidence supported the viewpoint that high CHGA expression is significantly associated with poor OS in PCa. Future studies with large sample sizes and stratified analyses according to clinicopathological characteristics are needed to explore tailored treatments that directly target CHGA for the improvement of survival in men with PCa.