Scolaris Content Display Scolaris Content Display

PARP (Poly ADP‐Ribose Polymerase) inhibitors for locally advanced or metastatic breast cancer

Collapse all Expand all

Background

Locally advanced and metastatic breast cancer remains a challenge to treat. With emerging study results, it is important to interpret the available clinical data and apply the evidence offering the most effective treatment to the right patient. Poly(ADP Ribose) Polymerase (PARP) inhibitors are a new class of drug and their role in the treatment of locally advanced and metastatic breast cancer is being established.

Objectives

To determine the efficacy, safety profile, and potential harms of Poly(ADP‐Ribose) Polymerase (PARP) inhibitors in the treatment of patients with locally advanced or metastatic breast cancer. The primary outcome of interest was overall survival; secondary outcomes included progression‐free survival, tumour response rate, quality of life, and adverse events.

Search methods

On 8 June 2020, we searched the Cochrane Breast Cancer Group Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE via OvidSP, Embase via OvidSP, World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) search portal and ClinicalTrials.gov. We also searched proceedings from the major oncology conferences as well as scanned reference lists from eligible publications and contacted corresponding authors of trials for further information, where needed.

Selection criteria

We included randomised controlled trials on participants with locally advanced or metastatic breast cancer comparing 1) chemotherapy in combination with PARP inhibitors, compared to the same chemotherapy without PARP inhibitors or 2) treatment with PARP inhibitors, compared to treatment with other chemotherapy. We included studies that reported on our primary outcome of overall survival and secondary outcomes including progression‐free survival, tumour response rate, quality of life, and adverse events.

Data collection and analysis

We used standard methodological procedures defined by Cochrane. Summary statistics for the endpoints used hazard ratios (HR) with 95% confidence intervals (CI) for overall survival and progression‐free survival, and odds ratios (OR) for response rate (RR) and toxicity.

Main results

We identified 49 articles for qualitative synthesis, describing five randomised controlled trials that were included in the quantitative synthesis (meta‐analysis). A sixth trial was assessed as eligible but had ended prematurely and no data were available for inclusion in our meta‐analysis. Risk of bias was predominately low to unclear across all studies except in regards to performance bias (3/5 high risk) and detection bias for the outcomes of quality of life (2/2 high risk) and reporting of adverse events (3/5 high risk).
High‐certainty evidence shows there may be a small advantage in overall survival (HR 0.87, 95% CI 0.76 to 1.00; 4 studies; 1435 patients). High‐certainty evidence shows that PARP inhibitors offer an improvement in PFS in locally advanced/metastatic HER2‐negative, BRCA germline mutated breast cancer patients (HR 0.63, 95% CI 0.56 to 0.71; 5 studies; 1474 patients). There was no statistical heterogeneity for these outcomes. Subgroup analyses for PFS outcomes based on trial level data were performed for triple‐negative breast cancer, hormone‐positive and/or HER2‐positive breast cancer, BRCA1 and BRCA2 germline mutations, and patients who had received prior chemotherapy for advanced breast cancer or not. The subgroup analyses showed a persistent PFS benefit regardless of the subgroup chosen. Pooled analysis shows PARP inhibitors likely result in a moderate improvement in tumour response rate compared to other treatment arms (66.9% vs 48.9%; RR 1.39, 95% CI 1.24 to 1.54; 5 studies; 1185 participants; moderate‐certainty evidence).
The most common adverse events reported across all five studies included neutropenia, anaemia and fatigue. Grade 3 or higher adverse events probably occur no less frequently in patients receiving PARP inhibitors (59.4% for PARP arm versus 64.5% for non‐PARP arm, RR 0.98, 95% CI 0.91 to 1.04; 5 studies; 1443 participants; moderate‐certainty evidence).
Only two studies reported quality of life outcomes so this was not amenable to meta‐analysis. However, both studies that did assess quality of life showed PARP inhibitors were superior compared to physician’s choice of chemotherapy in terms of participant‐reported outcomes.

Authors' conclusions

In people with locally advanced or metastatic HER2‐negative, BRCA germline mutated breast cancer, PARP inhibitors offer an improvement in progression‐free survival, and likely improve overall survival and tumour response rates. This systematic review provides evidence supporting the use of PARP inhibitors as part of the therapeutic strategy for breast cancer patients in this subgroup. The toxicity profile for PARP inhibitors is probably no worse than chemotherapy but more information is required regarding quality of life outcomes, highlighting the importance of collecting such data in future studies. Future studies should also be powered to detect clinically important differences in overall survival and could focus on the role of PARP inhibitors in other relevant breast cancer populations, including HER2‐positive, BRCA‐negative/homologous recombination repair‐deficient and Programmed Death‐Ligand 1 (PDL1) positive.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

PARP inhibitors for locally advanced or metastatic breast cancer

What is the aim of this review?

PARP inhibitors are a new class of drug that block DNA repair in tumour cells and hence lead to cell death. The aim of this Cochrane Review was to find out their efficacy and safety in the treatment of patients with locally advanced or metastatic breast cancer. The Cochrane review authors collected and analysed all relevant trials to answer this question and found five trials addressing this topic. Our primary aim was to look at whether PARP inhibitors prolonged survival. We also evaluated whether these drugs prolonged the time before disease progression (usually defined as growth of more than 20% or development of a new metastasis), caused the tumour to shrink, or resulted in more side effects.

What was studied in the review?

We included randomised controlled trials of participants with locally advanced or metastatic breast cancer comparing 1) chemotherapy in combination with PARP inhibitors, compared to the same chemotherapy without PARP inhibitors or 2) treatment with PARP inhibitors, compared to treatment with other chemotherapy. We included trials that reported on our primary outcome of overall survival and secondary outcomes including progression‐free survival, tumour shrinkage rate, quality of life, and side effects.

How up‐to‐date is this review?

We searched for published trials up to June 2020 and included the results of five trials involving 1474 participants.

What are the main results of the review?

For people with locally advanced or metastatic HER2‐negative (people with breast cancer that tests negative for a protein called human epidermal growth factor receptor 2), BRCA germline mutated (participant carries a mutation in the BRCA gene) breast cancer, our systematic review found that PARP inhibitors:

‐ may reduce the risk of death by 13% (i.e. people treated with these drugs live longer overall compared to those treated in the comparator treatment arm);
‐ reduce the risk of disease growth by 37%;
‐ may improve the chance of tumour shrinkage (66.9% for PARP inhibitors versus 48.9% for other treatments):
‐ result in little to no difference in side effects compared to other treatment arms.

Quality of life data were collected in two trials and the evidence available showed PARP inhibitors were superior compared to physician’s choice of chemotherapy in terms of participant‐reported outcomes.

Authors' conclusions

Implications for practice

Our systematic review shows PARP inhibitors offer a progression‐free survival advantage and there might be a small advantage in overall survival for patients with locally advanced/metastatic HER2‐negative, BRCA germline mutated breast cancer. Given PARP inhibitors did not have significantly increased toxicity and two studies that did look at quality of life outcomes showed an improvement, PARP inhibitors could be incorporated into the treatment paradigm for this subgroup of breast cancer patients.

However, the optimum method of use of PARP inhibitors remains unclear. The small numbers of trials were insufficient for meta‐analysis to determine whether they should be used as a single agent or in combination with chemotherapy.

Implications for research

Our systematic review has raised a number of unanswered questions and helps guide direction for future studies. Certainly, future studies with sufficient power to detect meaningful improvements in overall survival are needed.

HER2‐positive and BRCA‐negative/homologous recombination repair‐deficient populations were excluded from all studies, so further research with these populations is needed. Given the benefit seen in the metastatic setting, PARP inhibitors may also have a role in the neoadjuvant or the adjuvant settings.

Data to help clinicians with sequencing of therapies is lacking and further research in this area is vital to help answer when are PARP inhibitors most effective and whether they should be given alone or in combination. We believe further investigation into the role of platinum chemotherapy in this group is needed and whether PARP inhibitors are superior for patients who have had prior platinum exposure. For the subgroup of BRCA‐mutated PDL1‐positive patients, especially those with triple‐negative disease, it remains unclear as to whether immunotherapy or PARP inhibitors may be most beneficial. Also, it is unclear whether previous immunotherapy impacts PARP inhibitor utility.

Future studies need to collect data on quality of life to investigate the implications of using PARP inhibitors, especially in the palliative setting for patients with advanced breast cancer. In our current environment of scarce resources and a stressed health care system, it becomes even more apparent that assessing cost‐effectiveness of new interventions, in both health and economic terms, is essential.

Summary of findings

Open in table viewer
Summary of findings 1. PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer

PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer

Patient or population: locally advanced or metastatic breast cancer
Setting:
Intervention: PARPi‐containing regimen
Comparison: non‐PARPi regimen

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐PARPi regimen

Risk with PARPi‐containing regimen

Overall Survival**
follow up: 24 months

Study population

HR 0.84
(0.76 to 1.00)

1435
(4 RCTs)

⊕⊕⊕⊕
HIGH 1 2 3 4

550 per 1,000

497 per 1,000
(446 to 550)

Progression Free Survival**
follow up: 12 months

Study population

HR 0.63
(0.56 to 0.71)

1474
(5 RCTs)

⊕⊕⊕⊕
HIGH 1 3 5 6

625 per 1,000

461 per 1,000
(423 to 502)

Response Rate

Study population

RR 1.39
(1.24 to 1.54)

1185
(5 RCTs)

⊕⊕⊝⊝
LOW 1 3 6 7

489 per 1,000

695 per 1,000
(636 to 749)

Grade ≥3 AEs

Study population

RR 0.98
(0.91 to 1.04)

1443
(5 RCTs)

⊕⊕⊕⊝
MODERATE 1 3 8 9

645 per 1,000

620 per 1,000
(555 to 684)

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

** Given i) overall survival and progression‐free survival are continuous endpoints in clinical practice but ii) that continuous measures are not easily quantifiable (even if the HR is available), we opted to estimate the percentage of patients with this outcome (e.g. death) at a predefined time interval to practically estimate the size of treatment benefit for readers. We extrapolated this information from Kaplan‐Meier curves from the included studies. We started with the OS at 2 years, then subtracted this from 1 to arrive at incidence of death at 2 years and similarly for PFS at 1 year (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD).

CI: Confidence interval; HR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 All studies mostly graded as low to unclear risk of bias. This is based on the scores from each domain including 3/5 studies which had high risk of bias in terms of performance bias due to being open‐label. Also, detection bias for adverse events (3/5 studies) were judged as having high risk of bias. Overall, judged as unclear but not serious risk of bias.

2 I2=0%, indicating low heterogeneity.

3 No indirectness present.

4 95% CI did not extend past HR of 1.0 and the total number of patients exceeded 400.

5 I2=2%, indicating low heterogeneity.

6 95% CI excluded a HR of 1.0 and the total number of events exceeded 400.

7 Significant heterogeneity (I2=90%) without an obvious clinical explanation arising from differences in included trials.

8 Significant heterogeneity (I2=73%).

9 95% CI crosses both 1 (the point of no effect) and 0.75 (the point of significantly reduced toxicity)

Background

Description of the condition

Treatment of metastatic breast cancer remains a challenge to current patients and clinicians, with all patients eventually succumbing to the disease. Prognosis and survival rates vary greatly depending on the extent of the disease, performance status of the patients and the pathological tumour subtype, in particular, its immunohistochemical receptor status i.e. oestrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2). Expression of the ER and PR in tumours confers a better prognosis, while HER2‐positive and triple‐negative breast cancer (tumours without expression of ER/PR and HER2) tend to be indicative of a more aggressive cancer (Foulkes 2010). Breast cancer phenotypes have also been described in terms of tumour gene expression profiles which have approximate immunohistochemical correlates (Perou 2000). Basal‐like breast cancer is one such group and has similar morphological and genetic features with triple‐negative breast cancer, but they are not completely identical.

Breast cancers arising in patients with germline mutations in the BReast CAncer gene 1 (BRCA1) are more often triple‐negative and basal‐like, whereas BReast CAncer gene 2 (BRCA2)‐associated tumours are difficult to distinguish from sporadic cancers using standard histology techniques (Foulkes 2010).

Triple‐negative breast cancer and basal‐like breast cancer occur most frequently in young women and respond to conventional chemotherapy but relapse earlier and more frequently than hormone receptor‐positive breast cancer and are likely to have a less favourable overall outcome (Foulkes 2010).

Description of the intervention

Novel therapeutic strategies for metastatic breast cancer are in clinical development. Poly ADP Ribose Polymerase (PARP) inhibitors are one new class of agents. PARP‐1 and PARP‐2 proteins are part of the complex that is assembled in response to single‐strand deoxyribonucleic acid (DNA) breaks and are integral to the repair of single‐strand DNA breaks (Khasraw 2011). DNA damage induction is a common mode of action of many anti‐cancer drugs.

BRCA1 and BRCA2 are part of the complex that permits homologous recombination, i.e. repairing of a double‐strand DNA break. If these genes are mutated, then DNA cannot be repaired via homologous recombination (Davar 2012). PARP inhibitors promote the progression of single‐strand DNA breaks to double‐strand DNA breaks and can induce synthetic lethality in cells with impaired homologous recombination mechanisms, such as those with a BRCA mutation (Davar 2012).

Clinical trials of PARP inhibitors (PARPi) are ongoing in BRCA mutation‐associated cancers as well as sporadic breast cancers, ovarian cancers and other malignancies.

How the intervention might work

Ongoing studies have shed light on important genetic abnormalities in triple‐negative breast cancer, basal‐like breast cancer, and BRCA‐associated breast tumours. Treatment with PARP inhibitors has revealed encouraging data in early phase clinical studies in metastatic breast cancer including patients with triple‐negative breast cancer, basal‐like breast cancer and BRCA1‐associated tumours (Khasraw 2011). There are several ongoing larger studies using PARP inhibitors in breast cancer either as a single agent or in combination with other cytotoxic medications.

Why it is important to do this review

With emerging study results, it is important to interpret the available clinical data and apply the evidence offering the most effective treatment to the right patient. It is hoped that these agents will play a significant role in the treatment of cancers including those arising in BRCA1/2 mutation carriers. The work that has been done so far raises the possibility that future studies will uncover additional relationships between PARP‐dependent pathways and tumour‐specific defects present in sporadic cancers (Khasraw 2011). The optimal PARP inhibitor‐chemotherapy drug combination remains to be established, with a wide range of ongoing trials exploring these questions.

Objectives

To assess the effects of PARP inhibitors for patients with locally advanced or metastatic breast cancer.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs). We only included RCTs when the agent evaluated was mechanistically described as a PARP inhibitor, used alone or in combination with chemotherapy or other biologic agents (or a combination of these treatment modalities).

Types of participants

Patients with locally advanced or metastatic breast cancer.

Types of interventions

The intervention was the use of PARP inhibitors for locally advanced or metastatic breast cancer treatment. The comparator involved treatment with chemotherapy without PARP inhibitors.

We included trials involving:

  • chemotherapy in combination with PARP inhibitors, compared to the same chemotherapy without PARP inhibitors;

  • treatment with PARP inhibitors, compared to treatment with other chemotherapy

Types of outcome measures

Primary outcomes

  • Overall survival (OS), defined as the length of time from randomisation to death from any cause.

Secondary outcomes

  • Progression‐free survival (PFS), defined as the time from randomisation to either death or disease progression, whichever occurs first:

    • Time To Progression (TTP) used if PFS was not reported

  • Disease progression, as defined according to the widely‐used 'Response Evaluation Criteria In Solid Tumors (RECIST)' criteria used for solid tumours (Therasse 2000):

    • Other response criteria also accepted if they are well defined in the study;

    • Some currently‐open studies have abandoned RECIST scoring as the only method of assessing response and/or eligibility criteria partly due to recruitment issues, but also because it seems that it is not appropriate to assess response in some instances. This is similar to problems assessing activity of the biological (non‐cytotoxic) agents.

  • Quality of life (QoL), if a sufficient number of studies with adequate quality of life assessment are or become available and measured using a validated instrument e.g. the 36‐Item Short Form Health Survey (SF‐36), Functional Assessment of Cancer (FACT), the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life questionnaires

  • Adverse events, classified according to the World Health Organization (WHO) or National Cancer Institute‐Common Terminology Criteria (NCI‐CTC), including the percentage of treatment‐related deaths.

Search methods for identification of studies

Electronic searches

We conducted systematic literature searches to identify published and unpublished RCTs. Due to the relatively recent availability of PARP inhibitors, the start date for the literature search was 2008, which was considered sufficient for the purpose of this review. We identified RCTs using the following key words: including but not restricted to PARP inhibitors (specific drugs: veliparib, olaparib), and breast cancer.

We did not apply any language restriction in the searches.

We searched the following databases on 18 June 2020:

  1. the Cochrane Breast Cancer Group Specialised Register;

  2. the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library) (Appendix 1);

  3. MEDLINE via OvidSP (Appendix 2);

  4. Embase via OvidSP (Appendix 3);

  5. WHO International Clinical Trials Registry Platform (ICTRP) search portal for all prospectively registered and ongoing trials (http://apps.who.int/trialsearch/) (Appendix 4);

  6. ClinicalTrials.gov (http://clinicaltrials.gov/) (Appendix 5).

Searching other resources

We screened reference lists from trial publications selected by electronic searching in order to identify further relevant trials. We also searched published conference abstracts from the following organisations:

  • European Society for Medical Oncology (published in the Annals of Oncology);

  • European Council for Clinical Oncology (published in the European Journal of Cancer);

  • San Antonio Breast Cancer Symposium of the American Association of Cancer Research;

  • St. Gallen International Breast Cancer Conference;

  • American Society for Clinical Oncology;

  • American Society for Clinical Oncology Breast Cancer Symposium.

The search strategy was constructed using a combination of subject headings and text words relating to PARP inhibitors in breast cancer. In addition, we contacted members of the relevant cancer research groups, experts in the field, and manufacturers of relevant drugs for details of outstanding clinical trials and any relevant unpublished material. One conference abstract was provided by the editors.

Data collection and analysis

Selection of studies

Three review authors (AT, DC, MT) independently assessed the titles and abstracts retrieved by the search strategy for potential eligibility. We obtained full‐text articles of potentially eligible studies for further assessment (refer to Types of interventions). The three authors assessed these full‐text articles for risk of bias independently and in a blinded fashion (to authors, journal, drug company, institutions and results), with disagreement resolved by consensus with a fourth review author (MK).

We included abstracts or unpublished data only if sufficient information on the study design, characteristics of participants, interventions and outcomes was available. We attempted to obtain further information or final results from the primary trial author.

Data extraction and management

Two authors (AT, DC) performed data extraction independently. We entered data into the Cochrane Collaboration statistical software, Review Manager 2014. For each eligible trial, we recorded the following study characteristics: study design, participants, setting, interventions, 'Risk of bias' information, duration of follow‐up, efficacy outcomes, biomarker analyses and side‐effects.

For studies with more than one publication, we extracted data on all outcomes from the most relevant publication.

Two review authors (AT, DC) independently extracted details of study population, interventions and outcomes by using a standardised data extraction form. We resolved differences in data extraction by consensus with a third author (MK), referring back to the original article.

Our data extraction form included at least the following items.

  • General information: title, authors, source, contact address, country, published/unpublished, language and year of publication, sponsoring of trial.

  • Trial characteristics: study design, duration/follow‐up, 'Risk of bias' assessment as specified above.

  • Patients: inclusion and exclusion criteria, sample size, baseline characteristics, similarity of groups at baseline, withdrawals and losses to follow‐up.

  • Interventions: dose, route and timing of chemotherapy, PARP inhibitors therapy and comparison intervention.

  • Outcomes: hazard ratio (HRs) and 95% confidence intervals (CIs), log rank chi square, log rank P values, number of events, number of patients per group, median (one‐, two‐, three‐ and five‐year survival rates).

Assessment of risk of bias in included studies

Two independent authors (AT, DC) assessed all studies that met the inclusion criteria for risk of bias, with disagreement resolved by the third review author (MK). We assessed the risk of bias for every included study using the Cochrane Collaboration's 'Risk of bias' tool (Higgins 2011). The Cochrane Collaboration's tool for assessing risk of bias encompasses seven domains, and our judgements on these domains for each trial will be reported in the 'Risk of bias' table. The seven domains are:

  1. sequence generation;

  2. allocation concealment;

  3. blinding of participants, personnel;

  4. blinding of outcome assessment (assessed separately for outcomes overall survival, progression‐free survival & overall response rates, quality of life and adverse events);

  5. incomplete outcome data;

  6. selective outcome reporting (trial protocols and trial result publications will be cross‐checked); and

  7. other sources of bias (role of funding body considered).

In addition, we also included trials which permitted a cross‐over for patients after disease progression. However, in these trials, the number of patients who crossed over has to be considered in the interpretation of the results for overall survival. We considered the following criteria.

  • Was the allocation truly random?

  • Were groups similar at baseline regarding the most important prognostic factors?

  • Were the number of withdrawals, dropouts and losses to follow‐up in each group completely described?

  • Was the analysis done by intention‐to‐treat?

  • Were type and schedule of the follow‐up similar in the comparison group?

Measures of treatment effect

We performed meta‐analysis on the basis of published data; unpublished or updated data provided will be used, when available.

The summary statistics for time‐to‐event outcomes (i.e. overall survival and progression‐free survival) were hazard ratios (HRs) and their 95% confidence intervals (CIs) (Cox 1972). We estimated HRs and their 95% CIs directly or indirectly from the published data (Altman 2001). HRs can be estimated (under some assumptions) from log rank Chi2 values, from log rank P values, from observed to expected event ratios, from ratios of median survival times or time point survival rates (Machin 1997; Parmar 1998). For overall survival and progression‐free survival, we calculated a statistical summary of the results of all individual trials performed with one PARP inhibitor, that was used to compare the results for different drugs. We used the fixed‐effect model for meta‐analyses to compare treatment differences. We performed statistical analysis of summary data using Review Manager 2014.

For dichotomous outcomes (e.g. tumour response rates, adverse events), we expressed the treatment effect as a risk ratio (RR) with 95% CIs.

For continuous outcomes such as quality of life, we reported HRs comparing the time to achieve a clinically significant decrease in QoL.

We carried out intention‐to‐treat analyses for all outcomes.

Unit of analysis issues

There were no unit of analysis issues anticipated in this review.

Dealing with missing data

We contacted study authors to request missing data, where possible. Studies with insufficient data for a particular outcome were not included for analysis of that outcome.

Assessment of heterogeneity

We first inspected heterogeneity graphically using forest plots displaying effects of individual studies with 95% CIs. We also assessed heterogeneity of effects between studies using the Chi2 test and the degree of inconsistency among results of included studies using the I2 statistic (I2 value > 50% was considered as substantial heterogeneity and > 75% represented considerable heterogeneity) (Higgins 2011).

Assessment of reporting biases

We examined the possibility of publication bias using funnel plots as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Data synthesis

We extracted and entered the number of participants experiencing each outcome and total number of patients randomised to each study arm into Review Manager 2014 for statistical analysis. The data analysis adhered to the intention‐to‐treat principle.

For time‐to‐event outcomes, we conducted a fixed‐effect (inverse‐variance method) analysis, if appropriate. For dichotomous outcomes, we used the fixed‐effect model (Mantel‐Haenszel method) to calculate pooled results. For continuous outcomes, we conducted a fixed‐effect (inverse‐variance method) analysis, if appropriate.

We performed all analyses using Review Manager 2014 in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

We described the quality of the available evidence in 'Summary of findings' tables in line with recommendations from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We used the GRADEprofiler (GRADEpro) software to develop the tables (GRADEpro 2014). We selected four outcomes to include in the 'Summary of Findings' table: overall survival, progression‐free survival, tumour response rate and grade ≥ 3 adverse events.

Subgroup analysis and investigation of heterogeneity

The subgroups analysed included:

  1. Patients with triple‐negative breast cancer

  2. Patients with hormone‐positive and/or HER2‐positive breast cancer

  3. Patients with BRCA1 and BRCA2 germline mutations

  4. Patients who had received prior chemotherapy for advanced breast cancer or not

Our review was not able to report on subgroup analyses for OS outcomes as these data were not available for the included studies. Subgroup analyses for PFS outcomes was performed for triple‐negative breast cancer, hormone‐positive and/or HER2‐positive breast cancer, BRCA1 and BRCA2 germline mutations and patients who had received prior chemotherapy for advanced breast cancer or not.

Other potentially planned subgroup analyses in the protocol including lines of therapy, performance status, PARP inhibitors as monotherapy, maintenance therapy or in combination with chemotherapy (or chemo‐radiotherapy) and position in the breast cancer treatment paradigm (first line, second line or third line) and different PARP inhibitors, were not performed due to the lack of these data in the included studies (see “Differences between protocol and review”).

Sensitivity analysis

We planned to perform a sensitivity analysis for the primary outcome excluding studies at high risk of bias. However, all studies contributing to the OS analysis were judged as having low risk of bias so sensitivity analysis was not performed.

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach to assess the certainty of evidence of the key outcomes (Guyatt 2008). GRADEpro GDT software was used to develop the 'Summary of findings' table.

Two authors created the 'Summary of findings' tables using the following outcomes:

  • overall survival;

  • progression‐free survival;

  • tumour response rate;

  • grade 3 or higher adverse events.

Results

Description of studies

Results of the search

We identified 3547 references in the initial search and additionally 86 from trials registries, nine from handsearching and one was provided by the editors. After excluding 122 duplicates and 3406 nonrelevant and/or nonrandomised studies, we selected 115 records for further evaluation (Figure 1). We assessed these 115 records in full for eligibility. We excluded a further 11 that did not use PARP inhibitors, three that did not report PFS or OS, 19 nonrandomised studies, 26 neoadjuvant studies and six adjuvant studies.


Study flow diagram.

Study flow diagram.

One ongoing trial was identified and described in the Characteristics of ongoing studies table. Overall, six studies (49 records) were eligible but only five were included in the quantitative synthesis as no data were available from one study. Each of the five studies produced multiple publications (Figure 1).

Included studies

See Characteristics of included studies table.

Six studies were identified as eligible (BRAVO; BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD). Four were identified by the systematic search and two by handsearching. All six trials are multicentre randomised trials that included patients with advanced or metastatic breast cancer and most had a germline BRCA mutation.

The phase 3 BRAVO trial (BRAVO) of niraparib versus physician's choice in metastatic or locally advanced HER2‐negative, germline BRCA mutation‐positive breast cancer was set to determine whether niraparib had a similar beneficial effect in breast cancer as it has in ovarian cancer. This study ended prematurely because of poor patient adherence (OncologyPro 2020). The authors made multiple attempts to access any available data from this trial, however, none was available at time of publication. As such, this trial was not included in the meta‐analysis and risk of bias was not able to be assessed.

BROCADE 2 is a phase 2 partially blinded, multicentre, randomised controlled trial. Two hundred and ninety patients with advanced or metastatic breast cancer with a germline BRCA mutation and who had received two or fewer prior lines of therapy were randomised 1:1:1 to receive veliparib, carboplatin and paclitaxel (arm 1), placebo, carboplatin and paclitaxel (arm 2) or veliparib and temozolamide (arm 3). The third treatment arm of BROCADE 2 was not examined in any of our analyses, as it did not have a directly comparable control arm. The primary endpoint was PFS and secondary endpoints included OS, objective response rate, clinical benefit rate and safety.

BROCADE 3 is a phase 3 version of BROCADE 2. It is a double‐blind, placebo‐controlled, multicentre, randomised trial. Five hundred and thirteen patients with advanced or metastatic HER2‐negative breast cancer with a germline BRCA mutation who had received two or fewer prior lines of therapy were randomised 2:1 to receive veliparib, carboplatin and paclitaxel (arm 1) or placebo, carboplatin and paclitaxel (arm 2). The primary endpoint was PFS and secondary endpoints included OS, clinical benefit rate, objective response rate, and progression on subsequent therapy (PFS2).

EMBRACA is a phase 3 open‐label, multicentre, randomised controlled trial. Four hundred and thirty‐one patients with advanced or metastatic HER2‐negative breast cancer with a germline BRCA mutation who had received three or fewer prior lines of therapy were randomised 2:1 to talazoparib (arm 1) or physician’s choice of chemotherapy including capecitabine or eribulin or gemcitabine or vinorelbine (arm 2). The primary endpoint was PFS and secondary endpoints included OS, objective response rate, clinical benefit rate and safety.

Kummar 2016 is a phase 2 open label, multi‐centre, randomised, controlled trial. 45 patients with metastatic triple‐negative breast cancer who had received at least one prior line of therapy were randomised 1:1 to veliparib and cyclophosphamide (Arm 1) or cyclophosphamide alone (Arm 2). Primary endpoint was response rate and secondary endpoints included PFS and safety.

OLYMPIAD is a phase 3 open‐label, multicentre, randomised controlled trial. Three hundred and two patients with metastatic HER2‐negative breast cancer with a germline BRCA mutation who had received two or fewer prior lines of therapy were randomised 2:1 to olaparib (arm 1) or physician’s choice of chemotherapy including capecitabine or eribulin or vinorelbine (arm 2). The primary endpoint was PFS and secondary endpoints included OS, safety, time to second progression, death after first progression, objective response rate and health‐related quality of life.

As described, there were some differences in study design. OLYMPIAD and EMBRACA trials compared PARPi to physician's choice chemotherapy, whereas BROCADE 2, BROCADE 3 and Kummar 2016 compared PARP inhibitor in combination with chemotherapy to placebo plus chemotherapy. BROCADE 2 included HER2‐positive patients if they had progression on or were intolerant to HER2‐directed therapies, while all other studies excluded HER2‐positive patients.

Excluded studies

Three thousand five hundred and twenty‐one records were screened for eligibility. Three thousand four hundred and six were initially excluded as they were either nonrandomised or not relevant. One hundred and fifteen records were assessed in full for eligibility. Studies excluded included 11 that did not use PARP inhibitors, three that did not report PFS or OS and a further 19 that were nonrandomised. Studies examining PARP inhibitors in the neoadjuvant setting (26) and adjuvant setting (six) were also excluded and one ongoing trial was excluded. Below we have provided more detail regarding pertinent excluded studies.

The phase 2 and phase 3 studies by O'Shaughnessy and colleagues (O'Shaughnessy 2011; O'Shaughnessy 2014) investigated benefit for iniparib in patients with metastatic triple‐negative breast cancer by comparing iniparib plus gemcitabine and carboplatin versus gemcitabine and carboplatin alone. These studies were excluded as it was later discovered that iniparib is not a PARP inhibitor (Liu 2012).

The phase 2 ABRAZO trial (Turner 2019) investigated talazoparib in patients with locally advanced or metastatic breast cancer and a germline BRCA mutation with or without prior exposure to platinum agents. Response rates to talazoparib were higher in patients who had not had prior platinum exposure, suggesting some degree of cross‐resistance. This study was excluded as the intervention was prior exposure to platinum, not talazoparib.

The phase 2 RUBY trial (Patsouris 2019) reported clinical benefit in a subset of patients with germline BRCA wild‐type metastatic breast cancer whose tumour had high loss of heterozygosity scores. This study was excluded as it was a single‐arm study.

The phase 2 BRE09‐146 trial (Dwadasi 2014) reported that the addition of low dose rucaparib did not impact the toxicity of cisplatin or improve two‐year disease‐free survival in patients with BRCA mutations and/or triple‐negative breast cancer with residual disease after neoadjuvant therapy. The phase 3 randomised OlympiA trial (Tutt 2017) evaluated adjuvant use of olaparib in patients with high‐risk HER2‐negative breast cancer with germline BRCA mutations. These studies were excluded as they assessed PARP inhibition in the adjuvant setting.

Other trials of PARP inhibition in early stage breast cancer included the I‐SPY‐2 (Rugo 2016) and BrighTNess trials (Loibl 2018), which ultimately failed to show a benefit for adding the PARP inhibitor veliparib to standard neoadjuvant chemotherapy in patients with triple‐negative breast cancer. Both these studies were excluded from analysis as they looked at the role of PARP inhibitors in the neoadjuvant setting.
Investigators are continuing to work on translating the success of the included studies showing benefit of PARP inhibition in advanced breast cancer in the early breast cancer setting where ongoing trials incorporate new dosing schedules, PARP inhibitor monotherapy, and novel PARP combinations.

Risk of bias in included studies

Five out of six included studies (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD) were assessed with regard to risk of bias. BRAVO was deemed as having unclear risk of bias across all domains as the study ended prematurely and minimal details regarding the study were available. Figure 2 illustrates the 'Risk of bias' summary.


Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

EMBRACA was the only study to report central randomisation as a means of allocation concealment, making it low risk. The other four studies (BROCADE 2; BROCADE 3; Kummar 2016; OLYMPIAD) did not describe their methods of concealment. However, similar to the other studies, EMBRACA did not provide sufficient information about the sequence generation process to permit judgement meaning all five studies were judged as unclear risk of bias in regards to their sequence generation processes.

Blinding

In terms of performance bias, BROCADE 2 and BROCADE 3 were judged as having low risk as the studies were double‐blinded, where blinding of participants and key study personnel were ensured. OLYMPIAD, EMBRACA and Kummar 2016 were all open‐label studies, meaning there was no blinding of participants and personnel. So, these three studies were judged as having high risk of performance bias.

In terms of detection bias, individual outcomes were assessed for risk of bias. Overall survival is an objective measure and unaffected by lack of blinding so this outcome was judged as being at low risk for all five studies. BROCADE 2 and BROCADE 3 were double‐blinded studies so blinding of outcome assessment was guaranteed, meaning progression‐free survival and overall response rates outcomes were judged as being at low risk of bias. Similarily, EMBRACA and OLYMPIAD for progression‐free survival and overall response rate outcomes were judged as being at low risk of bias as radiological assessment was performed by blinded independent central reviews. Kummar 2016 had an unclear risk of bias for this domain as it was not stated whether radiological assessment was performed in a blinded fashion or not.

For the two studies that reported quality of life data (EMBRACA; OLYMPIAD), this outcome was judged as being at high risk of bias for both as both the studies were open‐label. In regards to adverse events, EMBRACA; Kummar 2016 and OLYMPIAD were judged as being at high risk of detection bias given the studies were open‐label. BROCADE 2 and BROCADE 3 had low risk of detection bias for adverse event reporting given that they were double‐blinded studies.

Incomplete outcome data

BROCADE 2 and OLYMPIAD were assessed as being at low risk since missing outcome data was balanced and small in numbers across intervention groups. Kummar 2016 was judged as being at unclear risk of attrition bias as six of the 45 enrolled patients (more than 10%) were not evaluable. Given BROCADE 3 has been published in abstract form only to date and participant attrition data were not provided by the lead author by time of submission, this was judged as being at unclear risk. EMBRACA reported 19 patients (18 in the standard therapy group and one in the talazoparib group) who withdrew consent; this imbalance of attrition between the two arms led to a judgement of unclear attrition bias risk.

Selective reporting

All five studies were judged as being at low risk of selective reporting bias because they reported on all prespecified endpoints in the identified study records.

Other potential sources of bias

No other potential sources of bias were identified.

Effects of interventions

See: Summary of findings 1 PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer

Five trials, involving 1474 patients, were included in this review (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD). Refer to summary of findings Table 1.

Primary outcomes

1.1 Overall survival

See: Figure 3; Analysis 1.1.


Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.1 Overall survival

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.1 Overall survival

Four of the five studies reported on OS (BROCADE 2; BROCADE 3; EMBRACA; OLYMPIAD). Pooled analysis of these trials with a sample size of 1435 patients showed PARP inhibitors may offer a small overall survival benefit, HR 0.87 (95% CI 0.76 to 1.00; P = 0.05; high‐certainty evidence), with no significant heterogeneity (I2 = 0%, P = 0.81).

Subgroup analyses for OS outcomes were not able to be performed as these data were not available for the included studies.

All studies contributing to the OS analysis were at low risk of bias so sensitivity analysis was not performed.

Cross‐over was only permitted in two trials (BROCADE 3; Kummar 2016), however, no hazard ratios were available regarding cross‐over analysis so no formal sensitivity analysis was done.

Secondary outcomes

1.2 Progression‐free survival

See: Figure 4; Analysis 1.2; Analysis 1.3; Analysis 1.4; Analysis 1.5; Analysis 1.6.


Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.2 Progression‐free survival

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.2 Progression‐free survival

Five of the trials reported PFS with a total of 1474 participants (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD). Pooled analysis of five trials with a sample size of 1474 patients showed PARP inhibitors prolong PFS, with a HR of 0.63 (95% CI 0.56 to 0.71; P < 0.00001; high‐certainty evidence), no significant heterogeneity (I2 = 2%, P = 0.39).

Subgroup analyses for PFS outcomes based on trial level data were performed for triple‐negative breast cancer, hormone‐positive and/or HER2‐positive breast cancer, BRCA1 and BRCA2 germline mutations and patients who had received prior chemotherapy for advanced breast cancer or not. Four trials were included for these pooled analyses (BROCADE 2; BROCADE 3; EMBRACA; OLYMPIAD). Kummar 2016 did not report subgroup outcomes.

All four trials included for subgroup analyses for PFS outcomes stipulated participants must carry a germline BRCA mutation as part of their inclusion criteria. For patients with BRCA1 mutated breast cancer (N = 717, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.65, 95% CI 0.53 to 0.78; P < 0.0001), with no significant heterogeneity (I2 = 0.0%, P = 0.62). For patients with BRCA2 mutated breast cancer (N = 697, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.62, 95% CI 0.51 to 0.76; P < 0.0001) with no significant heterogeneity (I2 = 0%, P = 0.42). There was no significant evidence of interaction between these two groups (subgroup interaction P = 0.78).

For patients with triple‐negative breast cancer (N = 664, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.61, 95% CI 0.47 to 0.80; P = 0.0003), with moderate heterogeneity (I2 = 44%, P = 0.15). Examination of the relevant studies revealed no particular explanation for this statistical heterogeneity, and thus no relevant subgroups of this subgroup analysis were further analysed.

For patients without triple‐negative breast cancer, that is, hormone‐positive and/or HER2‐positive breast cancer, (N = 771, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.66, 95% CI 0.53 to 0.82; P < 0.00001) with minimal heterogeneity (I2 = 23%, P = 0.27). There was no significant evidence of interaction between these two groups (subgroup interaction P = 0.68).

For patients who had received prior chemotherapy for advanced breast cancer (N = 729, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.64, 95% CI 0.53 to 0.77; P < 0.00001), with no significant heterogeneity (I2 = 0%, P = 0.57). For patients who had not received prior chemotherapy for advanced breast cancer (N = 706, 4 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.66, 95% CI 0.55 to 0.79; P < 0.00001), with no significant heterogeneity (I2 = 0%, P = 0.51). Further, there was no significant evidence of interaction between these two groups (subgroup interaction P = 0.82).

Although not prespecified, we also looked at PFS outcomes for patients who had received prior platinum chemotherapy compared to those who had not, as this is a clinically relevant distinction (see “Differences between protocol and review”). Three trials were included for this pooled analysis (BROCADE 3; EMBRACA; OLYMPIAD). For patients who had received prior platinum chemotherapy (N = 205, 3 RCTs), there was no significant evidence of PFS prolongation on pooling of studies (HR 0.71, 95% CI 0.50 to 1.01; P = 0.05), with no significant heterogeneity (I2 = 0%, P = 0.97). For patients who had not received prior platinum chemotherapy (N=1037, 3 RCTs), there was evidence of PFS benefit on pooling of studies (HR 0.63, 95% CI 0.53 to 0.73; P < 0.00001) with some heterogeneity (I2 = 16%, P = 0.30). There was no significant evidence of interaction between these two groups (subgroup interaction P = 0.52). These results should be interpreted with caution given lack of subgroup interaction and differences in subgroup sizes.

Other potentially planned subgroup analyses in the protocol, including lines of therapy, performance status, PARP inhibitors as monotherapy, maintenance therapy or in combination with chemotherapy (or chemo‐radiotherapy) and position in the breast cancer treatment paradigm (first‐line, second‐line or third‐line) and different PARP inhibitors, were not performed due to the lack of these data from the included studies (see “Differences between protocol and review”).

1.3 Tumour response rates

See: Figure 5; Analysis 1.7.


Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.3 Response rate

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.3 Response rate

Five trials reported overall response rates (that is, complete response plus partial response as per RECIST v1.1) in patients with measurable disease with a total of 1185 participants (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD). In two studies (OLYMPIAD; BROCADE 2), this was assessed by independent review and, in two studies (EMBRACA and BROCADE 3), this was assessed by investigators. Kummar 2016 did not specify the method of analysis. Pooled analysis shows PARP inhibitors likely improve response rates from 48.9% to 66.9% (RR 1.39, 95% CI 1.24 to 1.54; P < 0.00001; moderate‐certainty evidence). However, significant statistical heterogeneity was present (I2= 90%, P < 0.00001).

1.4 Quality of life

Two studies (EMBRACA and OLYMPIAD) reported quality of life outcomes so this was not amenable to meta‐analysis. OLYMPIAD and EMBRACA showed PARP inhibitors were superior compared to physician’s choice of chemotherapy in terms of patient‐reported outcomes. Both studies used the 30‐item EORTC Quality of Life Questionnaire (QLQ‐C30) to assess health‐related quality of life. In both studies, clinically meaningful deterioration was defined as a decrease of 10 points or more on the QLQ‐C30 and no subsequent observations with a decrease of less than 10 points from baseline. OLYMPIAD showed a small significant difference between treatment groups in the adjusted mean QLQ‐C30 score across all time points, and a clinically meaningful decrease in the QLQ‐ C30 score was delayed in the olaparib group (HR 0.44, 95% CI 0.25 to 0.77, P = 0.004). Similarly, EMBRACA showed significant overall improvements in global health status quality of life and significant delays in the times to clinically meaningful deterioration (HR 0.38, 95% CI 0.26 to 0.55, P < 0.0001).

1.5 Adverse events

See: Figure 6; Analysis 1.8; Analysis 1.9; Analysis 1.10; Analysis 1.11; Analysis 1.12.

All five trials reported on adverse events (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD). Pooled analysis of a sample size of 1443 patients showed likely little to no difference in rates of grade 3 or higher adverse events (59.4% for PARPi arm vs 64.5% for non‐PARPi arm, RR 0.98, 95% CI 0.91 to 1.04, P = 0.47; moderate‐certainty evidence). Significant statistical heterogeneity was present (I2= 73%, P = 0.005).


Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.4 Grade ≥ 3 AEs

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.4 Grade ≥ 3 AEs

The most common adverse events reported across all five studies included neutropenia, anaemia and fatigue. Neutropaenia was significantly less common in patients receiving PARP inhibitors (32.7% versus 52.0%, RR 0.67, 95% CI 0.60 to 0.76; P < 0.00001). Anaemia was more common in patients receiving PARP inhibitors (39.2% versus 33.7%, RR 1.21, 95% CI 1.04 to 1.41, P = 0.01). Fatigue was non‐significantly less common in patients receiving PARP inhibitors (32.0% versus 36.7%, RR 0.90, 95% CI 0.78 to 1.05, P = 0.18). Only four studies reported thrombocytopaenia event rates (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016). Thrombocytopaenia was non‐significantly less common in patients receiving PARP inhibitors (30.6% versus 35.8%, RR 0.98, 95% CI 0.84 to 1.15, P = 0.84).

Discussion

Summary of main results

We identified a total of 49 articles describing six eligible studies for qualitative analysis. Five studies comprising a total of 1474 participants were included in the meta‐analyses. Three studies were large phase 3 randomised controlled trials (BROCADE 3; EMBRACA; OLYMPIAD) and two were phase 2 trials (BROCADE 2; Kummar 2016). Trial design differed in that two of the phase 3 studies compared single agent PARP inhibitors to physician’s choice of chemotherapy (EMBRACA; OLYMPIAD) and the other three studies compared PARP inhibitors plus chemotherapy to placebo plus chemotherapy (BROCADE 2; BROCADE 3; Kummar 2016. Three studies used veliparib (BROCADE 2; BROCADE 3; Kummar 2016) and the others used olaparib (OLYMPIAD) and talazoparib (EMBRACA) in their PARP inhibitor intervention arms. Four of the five studies (BROCADE 2; BROCADE 3; EMBRACA; OLYMPIAD) were powered to assess progression‐free survival, with overall survival being a secondary outcome and one phase 2 study, the smallest, (Kummar 2016) assessed tumour response rate as the primary outcome and progression‐free survival as a secondary outcome.

The primary objective of this review was to assess the effects of PARP inhibitors for patients with locally advanced or metastatic breast cancer. Four out of five studies (1435/1474 patients) included only women with a germline BRCA mutation and similarly most participants were HER2‐negative (only 10 patients in BROCADE 2 were HER2‐positive) so our results pertain to these groups predominantly.

Our systematic review indicates that PARP inhibitors offer a PFS benefit in advanced/metastatic HER2‐negative, BRCA germline mutated breast cancer patients. Importantly, to the authors' knowledge, our systematic review provides the first evidence that PARP inhibitors may also offer a small advantage in overall survival. Pooled analysis also showed a likely improved response rate with PARP inhibitors. There was little to no heterogeneity in the overall analyses and most of the subgroup analyses, consistent with the observation that included studies were largely designed in a similar fashion. This provides evidence that the use of PARP inhibitors can form part of the therapeutic strategy for this subgroup of breast cancer patients. The risk of death at two years is 55% for patients not treated with PARPi, and 49.7% for patients treated with PARPi (equivalent to a number needed to treat for an additional beneficial outcome of 18.9). The risk of progression or death at one year was 62.5% for patients not treated with PARPi, and 46.1% for patients treated with PARPi (equivalent to a number needed to treat for an additional beneficial outcome of 6.1) (See summary of findings Table 1).

These positive overall findings are supported by the subgroup analyses all showing consistent findings of a persistent PFS benefit. We showed that PARP inhibitors confer a PFS benefit in advanced HER2‐negative, BRCA germline mutated breast cancer regardless of whether the patient carries a BRCA 1 or 2 mutation, whether the tumour is triple‐negative or not and regardless of whether chemotherapy was previously given in the metastatic setting.

Several studies have evaluated the use of platinum agents in patients with germline BRCA mutations. In particular, the TNT trial (Tutt 2018) showed an objective response rate of 68% with carboplatin versus 33% with docetaxel among 43 patients with metastatic triple‐negative breast cancer and a known BRCA mutation. That is, platinum chemotherapy is also an effective treatment for our patient population. Platinum‐based agents were not included as an option in the standard‐therapy groups of OLYMPIAD or EMBRACA so, given the results of the TNT study, the trials realistically only compared PARP inhibitors against second‐line therapies which is a limitation. It is unknown how PARP inhibitors would compare with first‐line drugs. In our systematic review, four out of the five studies (Kummar 2016 did not specify) excluded patients who failed prior treatment with platinum chemotherapy but patients were allowed prior platinum exposure in EMBRACA, OLYMPIAD and BROCADE 3. In this admittedly small subgroup of patients who had prior exposure to platinum‐based agents (205 patients), there was no significant PFS benefit demonstrated with PARP inhibitors, however, a 29% risk reduction was shown. As mentioned, these results need to be interpreted with caution and, given the small numbers, benefit from a PARP inhibitor in this particular subgroup should be confirmed in future studies. BROCADE 2 and BROCADE 3 assessed the efficacy of the combination of a PARP inhibitor and platinum chemotherapy acknowledging shared mechanisms of resistance and showed superiority with the combination compared to platinum chemotherapy alone. Head‐to‐head studies to compare platinum‐based agents with PARP inhibitors and to compare the response rates after progression among classes of inhibitors are lacking. A further limitation is the lack of taxane in the comparator arms of trials comparing PARP inhibitors to physicians' choice, such that it is unclear whether a patient should have a taxane or PARP inhibitor if taxane is being considered either first‐line (e.g. when platinum was given adjuvantly) or after platinum. This research is necessary to provide important insight into efficacy and sequencing.

Our systematic review indicates there is likely to be no increase in toxicity with PARP inhibitors when looking at overall grade 3 or above adverse events. Combining this with the added advantages of ease of administration and convenience of oral administration, PARP inhibitors are an appealing alternative to intravenous chemotherapy. It is important to note that our systematic review included studies using three different PARP inhibitors. Common toxicities for all PARP inhibitors included fatigue, gastrointestinal toxicities and cytopenias (Ho 2019, Murthy 2019). However, differences in toxicities between PARP inhibitors have emerged from the studies included in this review as well as other clinical trials (Murthy 2019). Olaparib’s most common grade 3 toxicity is anaemia (OLYMPIAD). The most frequent grade 3 or 4 toxicity for talazoparib was anaemia (EMBRACA). A phase 2 trial evaluating single agent veliparib in recurrent ovarian cancer patients showed the main grade 4 toxicity was thrombocytopenia (Coleman 2015). Our systematic review showed neutropenia was significantly less common and anaemia was significantly more common in patients receiving PARP inhibitors but there was not a meaningful difference in rates of other adverse events of any grade including thrombocytopaenia and fatigue. An upcoming systematic review and meta‐analysis may shed further light on the incidence of adverse events following use of different PARP inhibitors (Ho 2019).
Quality of life data were not easily available and were not amenable to quantitative analysis (only reported in two studies). This is obviously an important issue for both patients and clinicians, for objective quality of life indices help determine whether the adverse events from PARP inhibitors are outweighed by the known clinical benefit. This highlights the importance of collecting such data in future studies.

Overall completeness and applicability of evidence

The studies identified are relevant to the aims of this review and the clinical needs of the defined subgroup of patients with advanced breast cancer. This review allowed evaluation of all protocol‐defined endpoints, some fully and some partially. In terms of efficacy endpoints, OS, PFS, and response rates have been well reported with sufficient statistical information to allow meta‐analysis. On the other hand, documentation of quality of life outcomes was inadequate for meta‐analysis.

The overall survival results, our primary outcome, were limited by the lack of trials powered for overall survival as a primary endpoint. Also, only four out of five studies reported overall survival.

Our results were also limited by availability of only three phase 3 clinical trials, as well as the differing comparator arms of PARP inhibitor plus chemotherapy versus chemotherapy alone (BROCADE 2; BROCADE 3; Kummar 2016) compared to single agent PARP inhibitor versus physician’s choice of chemotherapy (EMBRACA; OLYMPIAD). Although there was minimal statistical heterogeneity, there is obvious clinical heterogeneity when combining outcomes from single‐agent PARP inhibitor studies and those from combination chemotherapy and PARP inhibitor studies. Potential ways to shed light on this heterogeneity would include between‐study analysis using individual patient data and/or a network meta‐analysis, however, data were not available to do this.

Our review identified studies that used three different PARP inhibitors. Given there were only five studies included, this means that there was a small number of studies for each PARP inhibitor. In pooling outcome data for the current study, we determined that there was an insufficient number of studies and no methodological grounds to perform analyses comparing the effects of individual PARP inhibitors. Therefore, investigation of which PARP inhibitor may be superior to another (and particularly in differing clinical settings) is outside the scope of the current study.

Baseline characteristics differed slightly in the trial populations. BROCADE 2; BROCADE 3 and EMBRACA included patients with locally advanced breast cancer whereas Kummar 2016 and OLYMPIAD only included metastatic patients. BROCADE 2 and BROCADE 3 had a higher proportion of patients without any prior chemotherapy treatment for metastatic disease compared to the other trials. BROCADE 2 is the only study to include HER2‐positive patients, albeit very small numbers (three in the PARP inhibitor arm and seven in the control arm).

In terms of applicability to current practice, our results support the use of PARP inhibitors in advanced/metastatic HER2‐negative, BRCA germline mutated breast cancer patients. This is in line with current FDA approvals. Olaparib was approved by the FDA in January 2018 for the treatment of patients with previously treated advanced or metastatic HER2‐negative, BRCA germline mutated breast cancer, on the basis of the progression‐free survival benefits published in the OlympiAD trial (OLYMPIAD). Talazoparib was approved for the same indication, based on the EMBRACA results, in October 2018. The third PARP inhibitor identified by our review, veliparib, has not yet been approved but the complete results of BROCADE 3 are yet to be formally published.

Quality of the evidence

Although only five studies were included for meta‐analysis and only three of these were phase 3 studies, there were a large number of patients (1474) and we believe that the conclusions drawn are clinically meaningful. We acknowledge the different choice of PARP inhibitors and difference in study design. However, there was homogeneity of results and minimal statistical heterogeneity, with each study individually concluding PARP inhibitors are superior to their comparator treatments and our meta‐analysis adds weight to this.

The results included in this systematic review allow a robust conclusion regarding the primary objective addressed in the overview, that of overall survival benefit in advanced/metastatic HER2‐negative, BRCA germline mutated breast cancer patients. Similar high‐quality evidence was available for the outcome of PFS. See summary of findings Table 1.

Lower quality evidence was available for the outcomes of response rates and grade 3 to 4 overall toxicity. We considered the quality of evidence for response rates to be low and was downgraded due to inconsistency. We considered the quality of evidence for grade 3 to 4 overall toxicity to be moderate and was downgraded due to imprecision. See summary of findings Table 1.

Potential biases in the review process

A thorough search was conducted to find all relevant studies including searching relevant databases and handsearching other resources. Relevant data were obtained from published sources, as well as an unpublished source in the case of BROCADE 3. We attempted to obtain complete data by contacting the authors, including data on overall survival from Kummar 2016. Methods used for the meta‐analysis were standard and unlikely to have introduced bias. We believe our systematic review and meta‐analysis has a low risk of bias.

Agreements and disagreements with other studies or reviews

There are a number of other meta‐analyses looking at PARP inhibitors in all tumour types but very few meta‐analyses looking at the role of PARP inhibitors in breast cancer specifically.

We note that Bao and colleagues (Bao 2016) published a meta‐analysis on the effectiveness and safety of PARP inhibitors in cancer therapy and concluded that PARP inhibitors do well in improving PFS with little toxicity, especially in patients with BRCA deficiency. However, no significant difference in overall survival between the PARP inhibitors and controls, even in the BRCA mutation group, was found. This meta‐analysis did not include the studies we have included for our meta‐analysis. In fact, the only studies that were included looking at PARP inhibitors in breast cancer specifically were two studies using iniparib (O'Shaughnessy 2011; O'Shaughnessy 2014), which has since been disproved to be a PARP inhibitor mechanistically (Liu 2012).

We note that Poggio and colleagues published a meta‐analysis on single‐agent PARP inhibitors for the treatment of patients with BRCA‐mutated HER 2‐negative metastatic breast cancer (Poggio 2018) and included two RCTs (EMBRACA; OLYMPIAD) for meta‐analysis. We agree with their conclusions that PARP inhibitors were associated with significantly improved PFS and response rates. They found no difference in overall survival when analysing only these two studies. Also, use of single‐agent PARP inhibitors was associated with a significantly increased risk of anaemia and any grade of headache, but a reduced risk of neutropenia and any grade of palmar‐plantar erythrodysesthesia syndrome, as compared with chemotherapy. This study has similar methodology to ours, however, we are investigating a larger group and therefore have relevant findings for more breast cancer patients.

There is growing evidence regarding the role of immune checkpoint inhibitors, especially in triple‐negative metastatic breast cancer. Impassion130 (Schmid 2018) showed a modest but statistically significant difference in progression‐free survival in favour of combination atezolizumab with nabpaclitaxel compared with nabpaclitaxel alone. However, the question remains whether PARP inhibitors or checkpoint inhibitors are more effective in metastatic BRCA mutated triple‐negative patients, or whether a combination may be superior.

From the current information available, including the addition of our systematic review results, it remains difficult to know which PARP inhibitor to choose for which patient. There are also two other PARP inhibitors on the market, niraparib and rucaparib, and these are under investigation for breast cancer patients. The phase 3 BRAVO trial, assessing single‐agent niraparib in locally advanced or metastatic, HER2‐negative, BRCA1/2‐mutated breast cancer, ended prematurely (BRAVO). But with more studies underway and the minimal chance of head‐to‐head studies being performed, it will likely come down to physician preference and drug availability.

Study flow diagram.

Figures and Tables -
Figure 1

Study flow diagram.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figures and Tables -
Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.1 Overall survival

Figures and Tables -
Figure 3

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.1 Overall survival

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.2 Progression‐free survival

Figures and Tables -
Figure 4

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.2 Progression‐free survival

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.3 Response rate

Figures and Tables -
Figure 5

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.3 Response rate

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.4 Grade ≥ 3 AEs

Figures and Tables -
Figure 6

Forest plot of comparison: 1 PARPi‐containing regimen vs non‐PARPi regimen, outcome: 1.4 Grade ≥ 3 AEs

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 1: Overall survival

Figures and Tables -
Analysis 1.1

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 1: Overall survival

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 2: Progression‐free survival

Figures and Tables -
Analysis 1.2

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 2: Progression‐free survival

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 3: Progression‐free survival: BRCA

Figures and Tables -
Analysis 1.3

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 3: Progression‐free survival: BRCA

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 4: Progression‐free survival: receptor status

Figures and Tables -
Analysis 1.4

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 4: Progression‐free survival: receptor status

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 5: Progression‐free survival: prior chemo for advanced breast cancer

Figures and Tables -
Analysis 1.5

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 5: Progression‐free survival: prior chemo for advanced breast cancer

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 6: Progression‐free survival: prior platinum exposure

Figures and Tables -
Analysis 1.6

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 6: Progression‐free survival: prior platinum exposure

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 7: Response rate

Figures and Tables -
Analysis 1.7

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 7: Response rate

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 8: Grade ≥ 3 adverse events

Figures and Tables -
Analysis 1.8

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 8: Grade ≥ 3 adverse events

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 9: Neutropenia (any grade)

Figures and Tables -
Analysis 1.9

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 9: Neutropenia (any grade)

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 10: Anaemia (any grade)

Figures and Tables -
Analysis 1.10

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 10: Anaemia (any grade)

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 11: Fatigue (any grade)

Figures and Tables -
Analysis 1.11

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 11: Fatigue (any grade)

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 12: Thrombocytopenia (any grade)

Figures and Tables -
Analysis 1.12

Comparison 1: PARPi‐containing regimen vs non‐PARPi regimen, Outcome 12: Thrombocytopenia (any grade)

Summary of findings 1. PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer

PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer

Patient or population: locally advanced or metastatic breast cancer
Setting:
Intervention: PARPi‐containing regimen
Comparison: non‐PARPi regimen

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐PARPi regimen

Risk with PARPi‐containing regimen

Overall Survival**
follow up: 24 months

Study population

HR 0.84
(0.76 to 1.00)

1435
(4 RCTs)

⊕⊕⊕⊕
HIGH 1 2 3 4

550 per 1,000

497 per 1,000
(446 to 550)

Progression Free Survival**
follow up: 12 months

Study population

HR 0.63
(0.56 to 0.71)

1474
(5 RCTs)

⊕⊕⊕⊕
HIGH 1 3 5 6

625 per 1,000

461 per 1,000
(423 to 502)

Response Rate

Study population

RR 1.39
(1.24 to 1.54)

1185
(5 RCTs)

⊕⊕⊝⊝
LOW 1 3 6 7

489 per 1,000

695 per 1,000
(636 to 749)

Grade ≥3 AEs

Study population

RR 0.98
(0.91 to 1.04)

1443
(5 RCTs)

⊕⊕⊕⊝
MODERATE 1 3 8 9

645 per 1,000

620 per 1,000
(555 to 684)

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

** Given i) overall survival and progression‐free survival are continuous endpoints in clinical practice but ii) that continuous measures are not easily quantifiable (even if the HR is available), we opted to estimate the percentage of patients with this outcome (e.g. death) at a predefined time interval to practically estimate the size of treatment benefit for readers. We extrapolated this information from Kaplan‐Meier curves from the included studies. We started with the OS at 2 years, then subtracted this from 1 to arrive at incidence of death at 2 years and similarly for PFS at 1 year (BROCADE 2; BROCADE 3; EMBRACA; Kummar 2016; OLYMPIAD).

CI: Confidence interval; HR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 All studies mostly graded as low to unclear risk of bias. This is based on the scores from each domain including 3/5 studies which had high risk of bias in terms of performance bias due to being open‐label. Also, detection bias for adverse events (3/5 studies) were judged as having high risk of bias. Overall, judged as unclear but not serious risk of bias.

2 I2=0%, indicating low heterogeneity.

3 No indirectness present.

4 95% CI did not extend past HR of 1.0 and the total number of patients exceeded 400.

5 I2=2%, indicating low heterogeneity.

6 95% CI excluded a HR of 1.0 and the total number of events exceeded 400.

7 Significant heterogeneity (I2=90%) without an obvious clinical explanation arising from differences in included trials.

8 Significant heterogeneity (I2=73%).

9 95% CI crosses both 1 (the point of no effect) and 0.75 (the point of significantly reduced toxicity)

Figures and Tables -
Summary of findings 1. PARPi‐containing regimen compared to non‐PARPi regimen for locally advanced or metastatic breast cancer
Comparison 1. PARPi‐containing regimen vs non‐PARPi regimen

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1.1 Overall survival Show forest plot

4

1435

Hazard Ratio (IV, Fixed, 95% CI)

0.87 [0.76, 1.00]

1.2 Progression‐free survival Show forest plot

5

1474

Hazard Ratio (IV, Fixed, 95% CI)

0.63 [0.56, 0.71]

1.3 Progression‐free survival: BRCA Show forest plot

4

1414

Hazard Ratio (IV, Fixed, 95% CI)

0.63 [0.55, 0.73]

1.3.1 BRCA 1

4

717

Hazard Ratio (IV, Fixed, 95% CI)

0.65 [0.53, 0.78]

1.3.2 BRCA 2

4

697

Hazard Ratio (IV, Fixed, 95% CI)

0.62 [0.51, 0.76]

1.4 Progression‐free survival: receptor status Show forest plot

4

1435

Hazard Ratio (IV, Random, 95% CI)

0.63 [0.54, 0.75]

1.4.1 Not triple negative

4

771

Hazard Ratio (IV, Random, 95% CI)

0.66 [0.53, 0.82]

1.4.2 Triple Negative

4

664

Hazard Ratio (IV, Random, 95% CI)

0.61 [0.47, 0.80]

1.5 Progression‐free survival: prior chemo for advanced breast cancer Show forest plot

4

1435

Hazard Ratio (IV, Fixed, 95% CI)

0.65 [0.57, 0.74]

1.5.1 Prior chemo

4

729

Hazard Ratio (IV, Fixed, 95% CI)

0.64 [0.53, 0.77]

1.5.2 No prior chemo

4

706

Hazard Ratio (IV, Fixed, 95% CI)

0.66 [0.55, 0.79]

1.6 Progression‐free survival: prior platinum exposure Show forest plot

3

1242

Hazard Ratio (IV, Fixed, 95% CI)

0.64 [0.55, 0.74]

1.6.1 Previous platinum

3

205

Hazard Ratio (IV, Fixed, 95% CI)

0.71 [0.50, 1.01]

1.6.2 No previous platinum

3

1037

Hazard Ratio (IV, Fixed, 95% CI)

0.63 [0.53, 0.73]

1.7 Response rate Show forest plot

5

1185

Risk Ratio (M‐H, Fixed, 95% CI)

1.39 [1.24, 1.54]

1.8 Grade ≥ 3 adverse events Show forest plot

5

1443

Risk Ratio (M‐H, Fixed, 95% CI)

0.98 [0.91, 1.04]

1.9 Neutropenia (any grade) Show forest plot

5

1443

Risk Ratio (M‐H, Fixed, 95% CI)

0.67 [0.60, 0.76]

1.10 Anaemia (any grade) Show forest plot

5

1443

Risk Ratio (M‐H, Fixed, 95% CI)

1.21 [1.04, 1.41]

1.11 Fatigue (any grade) Show forest plot

5

1443

Risk Ratio (M‐H, Fixed, 95% CI)

0.90 [0.78, 1.05]

1.12 Thrombocytopenia (any grade) Show forest plot

4

1147

Risk Ratio (M‐H, Fixed, 95% CI)

0.98 [0.84, 1.15]

Figures and Tables -
Comparison 1. PARPi‐containing regimen vs non‐PARPi regimen