Chronic sustained cell proliferation is one of the hallmarks of cancer, which is achieved through signalling changes resulting in progression through the cell cycle (1). Cyclin D1, along with its binding partners cyclin dependent kinases (CDK4/6), is a key regulator of the cell cycle, mediating transition from G₁ to S phase ( Figure 1 ). The gene encoding cyclin D1, CCND1, located on chromosome 11q13.3, has been reported to be amplified in 10-35% of breast cancers (BCa) (2–4) and its amplification has been associated with increased cyclin D1 expression (3). CCND1 amplification may be an effective prognostic and predictive biomarker in BCa.

Progression through the cell cycle is regulated by several cyclins and cyclin dependent kinases, at each stage of cycle. During the G₀ phase, retinoblastoma protein (Rb) inhibits the E2F transcription factor 1 (E2F1), thereby preventing transcription of G₁/S phase genes (5). The Cyclin D1-CDK4 or 6 complex phosphorylates the Rb, leading to disassociation of Rb from E2F1, thus activating G₁/S phase gene transcription and cell cycle progression (5) ( Figure 1 ). Growth factor signalling pathways, PI3K-AKT-mTOR and MAPK, are linked with both cyclin D1 and Estrogen Receptor (ER) activity (6, 7). There is interplay between the ER, CCND1 gene and cyclin D1 protein whereby the ER promotes transcription of the CCND1 gene, and the cyclin D1 protein interacts with the ER to promote ER mediated transcription (8–10).

The ER is expressed in approximately 75% of all BCa tumours (11). CCND1 amplification is particularly common in ER positive tumours and is associated with reduced survival in these patients (3, 12–15). Many ER positive patients are treated with endocrine therapies (ET), that inhibit the ER pathway, however up to 50% of patients develop resistance (16, 17). CCND1 amplification is a proposed mechanism of ET resistance, however this remains controversial due to studies with conflicting results. Some studies have shown significant association between CCND1 amplification and poor aromatase inhibitor (AI) (18, 19) and tamoxifen response (19), whilst another study has shown no association with tamoxifen response (20). Yet, another study suggested CCND1 was predictive of resistance to aromatase inhibitors but not to tamoxifen (21).

Added to the complexity of these biological relationships, is the role of cyclin D1 overexpression. Whilst the CCND1 gene amplification is detected in 10-35% of patients (2–4), 50-70% overexpress the cyclin D1 protein, suggesting additional mechanisms of cyclin D1 regulation (14). In a study that separated patients according to CCND1 amplification status and cyclin D1 overexpression; they found that patients with CCND1 amplification had reduced Recurrence Free Survival (RFS) compared to those with normal CCDN1, but patients with cyclin D1 overexpression had longer RFS than those without (22). Some studies have reported correlation between cyclin D1 expression and poor prognosis in ER positive BCa (23–25) but not all (26, 27). A meta-analysis concluded that cyclin D1 overexpression was a significant predictor of poor prognosis, in ER positive BCa but this was not observed in unselected BCa patients (25). The effects of cyclin D1 expression may differ due to differences in active signalling mechanisms between patients (26, 28).

The role of CCND1 amplification in BCa, in relation to ET response, RFS and overall survival (OS) remains unclear. This systematic review and meta-analysis evaluates the prognostic and predictive value of CCDN1 amplification in BCa patients across studies.

CCND1 is an oncogene that encodes the protein cyclin D1. Cyclin D1, in conjunction with CDK4/6, promotes progression through the cell cycle from G₁ to S phase. Amplification of CCND1 is a proposed marker of poor prognosis, in BCa, and in some studies is associated with ET resistance. Despite several studies investigating the predictive and prognostic value of CCND1 amplification, much ambiguity remains.

This meta-analysis included 18 studies comprising 6400 patients. Of these, 1136 (18%) had CCND1 amplified tumours, which is within the 10-35% range generally reported (2–4). Previous evidence suggested an association between CCND1 amplification and ER positive status (4, 13, 15, 51). Our analysis supports these findings, demonstrating a strong relationship between CCND1 amplification and positive ER status, with low heterogeneity. However, regarding other clinicopathological features, there were no significant relationships between CCND1 amplification and PR or HER2 status, nor with histological grade or tumour stage. Other studies have yielded conflicting results with respect to CCND1 amplification and PR status, HER2 status and histological grade (13, 15, 51–54). Many have reported no association between CCND1 amplification and tumour stage (4, 52–54). A previous meta-analysis found significant association between CCND1 amplification and ER and PR status as well as histologic grade, but no association with HER2 status or stage (4). Whilst this meta-analysis yielded similar results to one by He et al. (4), the included studies differed considerably; this is in part due to the search terms and exclusion criteria used, as well as several studies being published since their review. Our systematic review and metanalysis targeted studies of hormone receptor postmenopausal patients specifically with cohorts consisting of >35 patients and included analysis of CCND1 amplification and cyclin D1 overexpression.

There are several factors that may contribute to differences observed between studies, these may include variations in detection methods, cut off definitions for clinicopathological features and of CCND1 amplification (47), as well as the composition of cohorts in terms of molecular and histological subtypes (54, 55). In this study, analysis of CCND1 amplification in terms of molecular subtypes was not possible as it was not reported for many of the analysed studies, however others have shown CCND1 amplification is more common in the luminal subtype and associated with worse breast cancer specific OS compared to other subtypes (14, 56).

Our analysis also demonstrated that CCND1 amplification is strongly associated with high level of cyclin D1 expression. However, there was high heterogeneity between the studies, which may stem from the different methods of detection and scoring of immunohistochemistry results of cyclin D1 expression, as these were not standardised. Nevertheless, previous studies have noted the association between CCND1 amplification and high levels of cyclin D1 expression (13, 15, 54). Interestingly, one study showed that cyclin D1 immunostaining with an Allred score of >6.5 (57) was predictive of CCND1 amplification in ER positive BCa patients, with high sensitivity (94.2%) and specificity (87.8%) (13). This is somewhat surprising, given that CCND1 amplification occurs in 10-35% of BCa tumours ( Table 1 ), and cyclin D1 overexpression occurs in 50-70% of BCa tumours (2–4).

Another study found that CCND1 amplification and cyclin D1 overexpression correlated in ER positive but not ER negative BCa (58). This may be indicative of a positive feedback loop between cyclin D1 and ER, initiated by CCND1 amplification; cyclin D1 is known to stimulate ER transcriptional activity, whilst the ER forms a complex with Nuclear Factor Kappa B (NF-κB) and the cofactor, Rac Family Small GTPase 3 (RAC3) to promote CCND1 transcription (59). It has been found that high levels of Cyclin D1 mRNA was associated with positive ER status (24). Under this model, ER likely promotes expression of cyclin D1, which is further augmented by CCND1 amplification, accounting for the higher prevalence of cyclin D1 overexpression than amplification (22).

Regardless, CCND1 amplification retains its value as the preferred prognostic marker over cyclin D1 overexpression, due to several contradictory findings regarding the prognostic value of cyclin D1 mRNA and protein overexpression. There are several possible explanations for these discrepancies, i) differences between molecular subtypes, ii) treatment regime iii) mechanisms underlying cyclin D1 overexpression, iv) methodological differences. Because there is greater evidence for the CCND1 amplification as a prognostic maker, it has been suggested that studies are needed that consider patients with CCND1 amplification separately to those with cyclin D1 overexpression in the absence of amplification, in a well-defined cohort (60). Indeed, one study segregated patients into three groups; i) unamplified CCND1, cyclin D1 overexpression, ii) normal CCND1 and iii) CCND1 amplified and cyclin D overexpressed; they showed that patients with CCND1 amplification and cyclin D1 overexpression had worse RFS, whilst those with cyclin D1 overexpression but unamplified had good RFS, compared to those with normal CCND1 (22). This study demonstrated that there are important differences between these groups, that remain to be fully elucidated. Thus, the current evidence favours CCND1 amplification as the preferred prognostic marker over cyclin D1 transcript and protein expression. Additionally, our focus on CCDN1 amplification rather than transcript or protein is also based on practical considerations: with improving molecular technologies CCDN1 amplification testing could easily be moved into diagnostic settings as economic fast turnaround assay.

CCDN1 amplification is a proposed mechanism of resistance to ET. To examine this further, we analysed the effect of CCND1 amplification status on RFS and OS in ET treated patients. In these patients, CCND1 amplification was significantly associated with shorter RFS and OS. Removal of high RoB studies resulted in a stronger association between CCND1 amplification and RFS in ET treated patients. One of the limitations for these analyses is the number of studies which reported type of treatment with RFS and OS. The overall analysis for ET and RFS contained four studies, and this was further reduced to just two after exclusion of those assessed as having high RoB. Additionally, majority of studies in the ET RFS analysis reported results for treatment with tamoxifen alone, and hence may be biased towards tamoxifen treatment specifically. Comparison of different types of ET in terms of CCND1 amplification and RFS and OS, was not possible as majority of studies reported on tamoxifen only, and studies of other ETs were lacking. In fact, just one study (14) reported on ET generally, the others all focused on patients treated with tamoxifen. However, our findings agree with previous studies that reported CCND1 amplification as predictive of poor prognosis in patients treated with ET, including tamoxifen (15, 61) and AIs (12, 14, 19). However, these are not altogether unanimous; one study found that co-amplification of CCND1 and EMSY predicted tamoxifen resistance (62) whereas another found that CCND1 amplification was predictive of poor response to AIs but not tamoxifen (21). As mechanisms of ET action differ, it is possible that CCND1 amplification may contribute to resistance in some treatments but not others (63). However, differences may also be due to biological and chance differences between the cohorts, and further investigation is required to fully elucidate differences between subgroups of ET.

To explore the prognostic potential of CCND1 amplification, we compared duration of RFS and OS between patients with CCND1 amplified tumours and those without. In the analysis, including all eligible studies, we found that CCND1 amplification was significantly associated with shorter RFS but not OS. In both analyses there was a high degree of heterogeneity, and this could potentially be attributed to a range of variables. For example, RoB assessment indicated that multiple studies did not always report RFS and OS definitions or follow up times. Additionally, several studies failed to compare histological grade and tumour stage with CCND1 amplification. In support of this, removal of studies deemed as having a high RoB, yielded statistically significant association between CCND1 amplification and both RFS and OS, and reduced heterogeneity particularly in the case of OS.

There are some important limitations of the present study that should be considered. One of these is the variation amongst methods used to define CCND1 amplification and the cut off values. For example, many studies set cut offs based on a reference probe, with reference probes differing between studies, whilst others may select arbitrarily based on number of signals or the proportion of cells with positive signals. This has the potential to influence results in either direction depending on sensitivity of the assay, and on how cut offs were determined. Due to the wide variety of methods, and the overall number of studies included in this meta-analysis, it was not possible to conduct subgroup analyses between the different methods employed. Secondly, treatment regimens were generally poorly reported, making it difficult to compare amplification status with treatment responses. Thirdly, definitions of survival statistics differed in some studies, or were not reported. Fourthly, some studies only provided HR and 95% CI for subgroups of patients in which there was statistical significance; such underreporting of non-significant results, may have led to a bias in favour of associations between CCND1 and survival outcomes. Lastly, of the 18 studies included in our analysis, just three were rated as having low RoB across all six domains for quality assessment.

A major challenge with interpretation of data is due to the variation of methods used to detect CCND1 amplification, and variations in the cut-off values. Amplification detection methods include: southern blot, FISH, CISH, silver in situ hybridisation (SISH) PCR based (qPCR, quantitative fluorescence PCR, multiplex ligation-dependent probe amplification and droplet digital PCR), targeted sequencing, array comparative genomic hybridisation, and next generation sequencing (64). FISH was the most common detection method used in the literature, but there are potentially better methods. Increasingly, bright field ISH, CISH and SISH, are becoming the preferred method of amplification detection, owing to their increased resolution, and ability to simultaneously view gene amplification and tissue morphology (65). Currently, in the clinical setting, CISH and SISH are the preferred methods of HER2 amplification detection, with FISH used only in challenging cases (66). Current methods to define CCND1 amplification have not been taken up into routine diagnostic settings and development of better assays for rapid, reliable, economic, standardised CCND1 amplification testing are needed to harness the value of this prognostic and predictive biomarker.

In conclusion, our meta-analysis demonstrated that CCND1 amplification is significantly associated with positive ER status and cyclin D1 overexpression. CCND1 amplification was also predictive of both shorter RFS, and OS, in ET treated patients. As a prognostic biomarker, our meta-analysis indicated that CCND1 amplification may be effective in predicting shorter RFS and OS, after quality assessment. The lack of a standardised method of CCND1 amplification detection remains a considerable limitation, warranting future investigations aimed at establishing a standardised approach.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Conceptualisation, SJ; literature search and screening, SJ and BP; data collection SJ; risk of bias assessment, SJ and BP; Figures and tables, SJ; statistical analyses, SJ; drafting, SJ, BP, and TB; critical review of work, BP, SK, PS, HN, PdS, and TB. All authors contributed to the article and approved the submitted version.

This work was supported by a grant (13/TRC/1-01) from the Cancer Institute NSW through the CONCERT Translational Cancer Research Centre, SJ is a recipient of an Ingham Institute PhD Scholarship, generated by the Liverpool Catholic Club. BP is funded through a Clinical Academic Group Seed Grant from the Sydney Partnership for Health, Education, Research and Enterprise (SPHERE).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.