Introduction
In women, breast cancer is the most frequently diagnosed cancer worldwide, with an incidence of 1.7 million cases each year [
1]. Most cases, 75–80%, are hormone receptor positive, meaning that tumour cells express the oestrogen receptor (ER) and/or the progesterone receptor (PR). Curatively treated breast cancer patients are at risk of recurrence of disease. This occurs in approximately 10% of patients with hormone receptor-positive breast cancer within 5 years and continues to be a risk with an annual rate of 1.4–2.2% over more than 20 years [
2,
3]. Adjuvant systemic treatment diminishes the risk of recurrence, but can have adverse effects that negatively impact quality of life [
4].
The risk of recurrence in current clinical practice is estimated by considering classical prognostic factors, using nomograms such as the UK-based PREDICT tool or New Adjuvant Online [
5‐
7]. Despite the success of these risk prediction models to identify patients at high risk of recurrence based on clinical characteristics, prediction is on a population level and as a result leads to over- and undertreatment at a patient level [
8]. Prognostic biomarkers may improve the risk assessment, making it possible to better distinguish patients with a high risk of recurrence who may benefit from additional treatment, from patients with a low risk of recurrence for whom additional treatment may be omitted [
9]. This principle was recently demonstrated for both the Mammaprint and Oncotype DX biomarker assays by the MINDACT and TAILOR trials [
10,
11].
Biomarker research has increasingly incorporated epigenetic processes, particularly DNA methylation. DNA methylation is the addition of a methyl group to the carbon 5-position of cytosine within a cytosine guanine (CpG) dinucleotide. As methylation is a common and early event in cancer, and DNA methylation patterns differ between breast cancer molecular subtypes [
12,
13], alterations in the methylome form a potential class of biomarkers for early detection, prognosis and prediction to therapy [
14‐
16].
At the moment, DNA methylation markers are not yet being used in the clinical setting of breast cancer, despite the fact that many studies focused on the potential prognostic role of these markers and many DNA methylation markers have been suggested to have prognostic value [
17,
18]. Currently, an overview of these studies describing potential prognostic markers is lacking. In this systematic review, we provide a comprehensive overview of potential prognostic DNA methylation biomarkers for hormone-sensitive breast cancer. In addition, we comment on various methodological aspects of these biomarker studies, aiming to provide guidelines for optimising research into this subject.
Methods
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [
19]. No review protocol was previously published.
Eligibility criteria and study selection
Eligible articles were original research reports in the English language that had investigated hypo- or hypermethylated biomarkers in relation to patient survival or surrogate endpoints such as disease-free survival in breast cancer populations with oestrogen and/or progesterone receptor-positive breast cancer cases. We excluded in vitro studies, studies on non-human material, studies that focused on hereditary breast cancer cases, studies that focused on non-CpG DNA methylation and studies that had reported large amounts of data from biomarker arrays without further specification of the data to a single potential biomarker or biomarker panel.
Search strategy
PubMed and EMBASE were searched up to November 2018 for eligible studies using the following keywords and equivalents of these: ‘breast cancer’, ‘DNA based methylation biomarker’, ‘hormone receptor positive’ and ‘prognostic or predictive clinical outcome’ (see Additional file
1: Table S1 for a complete overview of the search terms).
Two reviewers (FH and TR) independently selected studies based on title, abstract and in selected cases full text. Disagreement was resolved by discussion between the reviewers until consensus was reached. References of selected studies were crosschecked for additional studies that were eligible for inclusion.
Data collection and extraction
The following data from all selected studies was collected independently by two reviewers (FH and TR): year of publication, study design, study population, length of follow-up, assay type and cut-off used, sequence of primer or probe, statistical methods used and reported association between marker and patient outcome were collected from all selected studies. When available, both univariate and multivariate outcome measures were collected. Study population information consisted of population size, country of patient selection, age, grade, hormone receptor status, HER2 status and stage according to the reported American Joint Committee on Cancer classification [
20]. The level of evidence (LOE) was assessed for each publication according to criteria as defined by Hayes et al. [
21] and the OCEBM Levels of Evidence Working Group [
22].
For each publication, all study endpoints on outcome were collected and compared with ‘the proposed Standardized definitions for Efficacy EndPoints in adjuvant breast cancer trials’ (STEEP) [
23]. Endpoints not defined in accordance to STEEP definitions were converted to STEEP-defined endpoints when sufficient information was provided. All defined biomarkers were checked for aliases in the NCBI Gene database and were reported by their current RefSeq gene names.
Analysis of reporting
All selected articles were scored according to the ‘REporting recommendations for tumour MARKer prognostic studies’ (REMARK) criteria [
24,
25]. The REMARK checklist consists of 20 items containing one or multiple sub-items. A single item was scored with one point if all relevant sub-items were reported, half a point if only part of the information was reported or zero points if no information on this item was reported. The REMARK checklist is presented in Additional file
2: Table S2. Scoring was performed by two independent researchers (FH and TR). If the total score per article differed, the differences were discussed until agreement was achieved.
REMARK scores were used to assess the risk of potential selection, measurement and confounding bias. The risk for selection bias was assessed by REMARK item #2 (‘patient characteristics’) and #6 (‘sample selection and follow-up’). Studies obtaining < 1.5 points for these combined items were considered to have an increased risk. Measurement bias regarding the assay method was assessed using REMARK items #5 (‘assay method’) and #11 (‘handling of marker values’). REMARK item #7 (‘clinical endpoint definition’) was employed to assess the risk of measurement bias regarding outcome assessment; incomplete or lack of reporting of this item (score < 1) was considered at risk for measurement bias. Confounding bias was assessed using REMARK criterion #16 (‘multivariable analysis’), as in multivariate analysis (score = 1) potential confounding is taken into account. In order to investigate the effect of study design on marker significance, we compared REMARK scores between studies that found significant results and studies that did not find significant results using a Wilcoxon signed-rank test.
Forest plots
A forest plot was prepared for all methylation markers that were investigated in two or more study populations. When included studies reported results for more than one location per marker or reported results derived from more than one source of DNA, such as primary tumour tissue or blood serum, all reported results were represented in the forest plot. If available, multivariable hazard ratios (HRs), 95% confidence intervals (CI) and p values were used. When studies reported only p values without HRs, these were still included in the forest plot, in order to give a complete overview. The statistical programming language R (version 3.3.1) was used to perform all analyses and generate the figures.
Discussion
In this systematic review, we provide an overview of prognostic DNA methylation markers for ER- and/or PR-positive breast cancer. We identified promoter hypermethylation of RASSF1, BRCA1, PITX2, CDH1, RARB, PCDH10 and PGR as well as the marker panel GSTP1, RASSF1 and RARB as possible markers of poor disease outcome. Four of these markers (RASSF1, PITX2, PCDH10 and the panel) were also shown to be of prognostic value independently of clinically relevant prognostic factors, suggesting that these markers may provide additional prognostic information. This may help to identify patients at increased risk of disease recurrence and to inform the choice of adjuvant therapy.
Although promising, current LOE for these markers is low, either level 3 or 4. Several explanations can be suggested for this low LOE. Most studies were performed retrospectively, which provides a lower LOE as compared to prospectively designed studies. To overcome this, biomarker research should preferably select patients from previously established prospective cohorts [
96]. In addition, only 18 markers and one marker panel were tested in multiple patient populations, and studies that did investigate the same marker showed extensive heterogeneity in technical assays, study endpoints and patient selection. This heterogeneity impaired comparison between studies and the performance of meta-analyses, making it impossible to combine low LOE studies in order to reach a higher LOE.
Heterogeneity between individual studies was introduced by several factors. DNA methylation can be analysed using several different techniques. Studies included in this systematic review applied nine different assays for determining methylation status. Although it has been shown that varying techniques could lead to different results [
97,
98], this is not always the case. In previous research, we have shown that the prognostic impact of a DNA methylation biomarker is not affected by the applied technique if the chosen technique is optimised correctly [
99]. Optimisation depends on correctly chosen cut-off values, assay conditions, origin and quality of the used source DNA and the location in which methylation is analysed [
99‐
101]. These factors all determine whether a sample is identified as methylated or unmethylated, directly influencing the sensitivity and specificity of the assay and should therefore be reported in great detail [
24,
25]. In our review, almost none of the included studies sufficiently reported these factors, as is also illustrated by a median REMARK score of 12. Recent research has shown the 5-hydroxy-methylation is a separate entity in epigenetic DNA alterations; however, as most currently applied techniques are incapable of discerning DNA methylation from 5-hydroxy-methylation, we have considered this distinction outside the scope of this review.
Apart from the chosen assay characteristics, heterogeneity in study endpoints was seen for the included studies. Although 85% of all studies reported the used endpoint, these endpoints were frequently not clearly described. Due to the long median survival in early breast cancer patients, overall survival is generally not feasible as an endpoint. Therefore, surrogate endpoints relating to disease recurrence are often applied. Recurrence in breast cancer can have many forms, such as locoregional recurrence, distant recurrence or second primary disease. As different types of recurrence are related to different patient, tumour and treatment characteristics, a precise definition of surrogate endpoints is needed [
23]. In addition, endpoint selection should be tailored to the envisioned purpose of the envisioned marker. For example, when a marker is studied with the goal of predicting the risk distant recurrence, distant recurrence-free survival or distant recurrence-free interval would include the most relevant events [
23].
Differences in tumour and treatment characteristics between studies were an additional source of heterogeneity. The treatment patients received, the percentage of patients that had hormone receptor-positive breast cancer or amplification of the HER2 gene differed markedly. Moreover, these characteristics, though vital for interpretation of the results of the studies, were often reported incompletely. The treatment regimen was only specified in 65% of the included studies. When treatment was specified, it was often described as ‘according to local guidelines’, which can vary per region, but also per time period. In breast cancer, the status and prognostic effect of biomarkers may change due to a specific treatment and it should therefore be considered when interpreting study results [
102]. The risk of breast cancer recurrence is directly correlated to the ER, PR and HER2 status [
5‐
7]. A lack of a detailed description of the study population makes it difficult to perform a meta-analysis or to identify a clinical setting in which a marker may be of use [
24,
25]. In addition, there was also a great variation in the covariates used in the multivariable analyses. In order to interpret the prognostic value of a marker, at least all currently used clinical prognostic factors, i.e. TNM classification, tumour grade, ER status, PR status and HER2 status, should be included [
24,
25]. Many studies did not perform these analyses or omitted key covariates without explanation.
The studies summarised in this review show numerous promising DNA methylation biomarkers for hormone receptor-positive breast cancer. Unfortunately, a meta-analysis of these studies is not possible due to the differences between the included studies. Additional research is needed to establish the prognostic value of these markers in predicting distant recurrence when used in addition to existing tests. Future research should be designed to prevent selection and confounding bias and should report findings in adherence to the REMARK criteria. In addition, measurement bias should be prevented by the usage of internationally accepted endpoints reported in the STEEP guidelines for breast cancer endpoint reporting [
23]. In order to get closer to clinical implementation, studies with a higher LOE are warranted. A feasible strategy may be to select patients from previously established prospective cohorts [
96].
In this review, we have not addressed the rational mechanistic pathways linking the investigated markers to breast cancer recurrence, as in many of the included studies this aspect is not explored. Functional exploration of epigenetic markers can help in marker validation as it adds a hint towards causation, which often lacks in observational epigenetic research [
18]. However, if a marker is thoroughly validated, it can be of clinical use without being mechanistically understood [
18]. We acknowledge that the REMARK criteria were designed as reporting guidelines and not as a tool for quality assessment. As reporting quality and study quality are not synonymous, the REMARK score is not a quality indicator as such, although we did find a relation between the REMARK score and reporting of statistically significant results. The REMARK score should not be regarded as a rating, but as a tool used to identify weaknesses in research. Some included studies analysed methylation as a side objective, rather than a main study objective, resulting in less well-described methodology and thus poor REMARK performance. A low REMARK score should therefore not be mistaken for an indicator of a poor marker, but rather an indication this marker needs further investigation.
Conclusion
In this systematic review, we provided a comprehensive overview of the available literature on prognostic DNA methylation biomarkers in ER- and/or PR-positive breast cancer. We identified hypermethylation of RASSF1, BRCA1, PITX2, CDH1, RARB, PGR, PCDH10 and a panel of GSTP1, RASSF1 and RARB as potential markers of poor disease outcome. We also provided an analysis of study reporting, which indicates high heterogeneity in currently published literature on this subject. Future prognostic DNA methylation marker research would benefit from standardised DNA methylation assessment methods, thorough study reporting and the use of standardised endpoint definitions.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.