Poly (ADP-ribose) polymerase inhibitors in solid tumours: Systematic review and meta-analysis

Francesco Schettini a,b,c,*, Fabiola Giudici d, Ottavia Bernocchi e, Marianna Sirico f, Silvia P. Corona e, Mario Giuliano a, Mariavittoria Locci g, Ida Paris h, Giovanni Scambia h,i,
Sabino De Placido a, Pasquale Rescigno j, Aleix Prat b,c,k,
Giuseppe Curigliano l, Daniele Generali e,f

a Department of Clinical Medicine and Surgery, University of Naples “Federico II”, Naples, Italy b Translational Genomics and Targeted Therapies in Solid Tumours, IDIBAPS, Barcelona, Spain c SOLTI Breast Cancer Research Group, Barcelona, Spain
d Unit of Biostatistics, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padova, Padova,
e Department of Medical, Surgery and Health Sciences, University of Trieste, Trieste, Italy
f Breast Cancer Unit, Azienda Socio Sanitaria Territoriale di Cremona, Cremona, Italy
g Department of Neuroscience, Reproductive Sciences and Dentistry, University of Naples Federico II, Naples, Italy
h Department of Woman and Child Health, Fondazione Policlinico Universitario A. Gemelli IRCCS, Roma, Italy
i Universita` Cattolica Del Sacro Cuore, Rome, Italy
j Interdisciplinary Group for Translational Research and Clinical Trials (GIRT-Uro), Candiolo Cancer Institute, FPO-
IRCCS, Candiolo, Italy
k Department of Oncology, Hospital Clinic of Barcelona, Barcelona, Spain
l Istituto Europeo di Oncologia, IRCCS ed Universita` di Milano, Milano, Italy

Received 11 December 2020; received in revised form 12 February 2021; accepted 22 February 2021
Available online 13 April 2021

Abstract Background: Poly (ADP-ribose) polymerase-inhibitors (PARPis) showed antitu- mour activity in BRCA1/2-mutated cancers, with more heterogeneous outcomes in tumours harbouring mutations that impair other genes involved in the DNA homologous recombina- tion repair (HRR) or wild-type (wt).
Methods: We conducted a systematic review and meta-analysis to better assess the role of PARPis in the treatment of metastatic solid tumours, with and without BRCA1/2 mutations. The primary end-point was progression-free survival (PFS). The secondary end-points were

* Corresponding author: Translational Genomics and Targeted Therapiesin Solid Tumours, August Pi I Sunyer Biomedical Research Institute (IDIBAPS), Carrer de Villaroel 170, Barcelona, 08036, Spain.
E-mail address: [email protected] (F. Schettini).

0959-8049/ª 2021 Elsevier Ltd. All rights reserved.

overall response rate (ORR) and overall survival (OS). A random-effects model was applied. Results: Twenty-nine studies (8,839 patients) were included. PFS was significantly improved (hazard ratio [HR]: 0.59, 95% confidence interval [CI]: 0.51e0.68, p < 0.001), without being affected by BRCA mutational status (p Z 0.65). Significant subgroup differences were observed with regard to the tumour site (p Z 0.001), line of therapy (p Z 0.002), control
arm (p < 0.001), type of PARPi (p < 0.001) and trials’ phase (p Z 0.006). PARPis were asso- ciated with ORR (relative risk: 1.35, 95% CI: 1.16e1.56, p < 0.001), with significant subgroup differences observed with regard to treatment line (p Z 0.03), control arm (p Z 0.04) and PARPis (p < 0.001) and independent of mutational status (p Z 0.44), tumour site (p Z 0.86) and trials’ phase (p Z 0.09). OS was significantly improved by PARPis (HR: 0.86, 95% CI: 0.80e0.92, p < 0.001), regardless of mutational status (p Z 0.57), tumour site (p Z 0.82), treatment line (p Z 0.22), control arm (p Z 0.21), PARPis (p Z 0.30) and trials’
phase (p Z 0.26). Finally, an exploratory subgroup analysis showed a significant PFS improvement (HR: 0.51, 95% CI: 0.43e0.60, p < 0.001) with PARPis in BRCA-wt/HRR- deficient tumours.
Conclusion: Our results confirm the efficacy of already approved PARPi-based treatments in BRCA1/2-mutant solid tumours, support their role also in BRCA-independent HRR-deficient tumours and suggest a potentially broader efficacy in some wt tumours, perhaps with appro- priate therapeutic partners. Prospective studies are warranted.
ª 2021 Elsevier Ltd. All rights reserved.

1. Introduction

Around 1 in 400e800 people harbours a germ line pathogenetic variant of BRCA1 and/or BRCA2 genes [1]. These genes are involved in the homologous recombination mechanism of repair (homologous recombination repair [HRR]) of DNA double-strand breaks (DSBs), a substantially error-free procedure [2]. Inactivation of BRCA1/2 due to pathogenetic mu- tations impairs HRR, thus indirectly inducing an incorrect processing of DSBs through inappropriate and error-prone alternative mechanisms of repair (i.e. non-homologous end joining and single-strand anneal- ing). This can lead either to a progressive accumulation of DNA lesions, which ultimately induce cell death via apoptosis [3,4], or to an increasing chromosomal instability, cell mutability and subsequent neoplastic transformation [2,5]. Indeed, inactivating germ line mutations in BRCA1 and 2 significantly increases the risk of early-onset breast cancer (45e65% lifetime risk) in both women and men and ovarian cancer (15e40%) [1,6]. Proportional to its prevalence, BRCA1 mutations also increase the risk of fallopian tube, peritoneal, testicular, prostate and pancreatic cancer [1,6,7]. Pathogenic variants in BRCA2 are also associ- ated with prostate and pancreatic cancer, as well as melanoma [1,6,7]. The risk of lung and gastric cancer is also slightly increased in patients with BRCA mutation [7]. Overall, the tumours with the highest prevalence of BRCA1/2 mutations are epithelial ovarian (10e15%), breast (2e10%), prostate (5e13%) and pancreatic cancer (4e7%) [8e11].

Different from BRCA1 and 2, poly (ADP-ribose) polymerase (PARP) is a family of nuclear enzymes involved in the recognition and repair of DNA single- strand breaks. The main activity of PARP is poly- ADP ribosylation (PARylation) of key components of chromatin and other proteins involved in DNA repair, as well as auto-PARylation. PARP1 is able to open up the chromatin, facilitating the recruitment of down- stream DNA repair factors [12]. Auto-PARylation triggers the release of bound PARP from DNA to allow the access of other DNA repair proteins [12]. As previously mentioned, in normal cells, DSBs are pri- marily repaired through HRR. However, when HRR is constitutionally dysfunctional (as in BRCA-mutant [mut] tumours), if other events that impair DNA dam- age repair occur, the damage is likely to become per- manent, with progressive accumulation of DNA lesions that ultimately leads cells to apoptosis [3,4]. This mechanism is on the basis of the theory of synthetic lethality, which justified the development of PARP in- hibitors (PARPis) for the treatment of BRCA-deficient tumours [13]. Two proof-of-concept phase II studies demonstrated the significant activity and good safety of the oral PARPi olaparib in metastatic breast cancer (MBC) and metastatic ovarian cancer (MOC), paving the way for its further development in these and other solid tumours, along with other PARPis, including talazoparib, niraparib, rucaparib and veliparib [7]. This drug class was originally considered to act through mere inhibition of PARP1/2 by competing with NADþ for the enzymes’ catalytic site (catalytic inhibition) [14,15]. On the contrary, PARPis have been recently demonstrated

136 F. Schettini et al. / European Journal of Cancer 149 (2021) 134e152

to elicit synthetic lethality in HRR-deficient cancers, mostly by inhibiting PARylation. This results in PARP molecules trapping on DNA, especially PARP1 (PARP trapping), avoiding further binding of other members of the DNA repair systems. In addition, aberrant PARP1- DNA complexes induce a stalling of the replication fork, leading to the generation of DNA DSBs, ultimately eliciting a cytotoxic effect [16].
Actually, this mechanism seems to be the most rele- vant contribution to synthetic lethality provided by PARPis, with different impact on both efficacy and toxicity [13,16]. Importantly, this mechanism of action might be also on the basis of a PARPi’s synergistic effect with other cytotoxic drugs, such as alkylating agents [16].
At present, there are more than 150 trials for multiple PARPis (e.g. niraparib, olaparib, rucaparib, talazoparib and veliparib) in different stages of development, com- bined with other drug classes or as a single agent [17]. Of those, at least 59 in several metastatic cancers have been published so far, including 29 phase II/III randomised controlled trials providing compelling evidence of effi- cacy in BRCA-mut tumours [7,8,18e49]. Nevertheless, there are still conflicting results with regard to PARPi efficacy in BRCAewild-type (wt) tumours, with or without deleterious mutations occurring in other HRR genes and depending on the PARPi administered. We thus performed a systematic literature review and meta- analysis to more precisely assess the role of PARPis in the treatment of metastatic solid tumours, with or without BRCA mutations.

2. Methods

2.1. Study objectives

The objective of our study was to comprehensively evaluate the activity and efficacy of PARPis in meta- static solid tumours, with or without BRCA1/2 muta- tions. The primary end-point was progression-free survival (PFS), whereas the secondary end-points were overall response rate (ORR) and overall survival (OS), as per the US Food and Drug Administration (FDA) guidance document [50].

2.2. Search strategy and selection criteria

After a systematic review of the literature was conducted in August 2020 on PubMed®, Cochrane Library and Embase®, we selected all phase II/III randomised clin- ical trials (RCTs) published until 31st July 2020 that studied the activity and/or efficacy of PARPis, combined or not combined with chemotherapy (CT) or other therapies, in metastatic solid tumours, independent of BRCA mutational status [8,18e48]. All other types of studies were excluded, including early-stage trials,

because different clinical settings imply different prog- nosis, therapeutic approaches (e.g. curative surgery, radiotherapy and so on) and end-points.
We used a query based on the words ‘parp inhibitors’, ‘niraparib’, ‘olaparib’, ‘talazoparib’, ‘veliparib’, ‘ruca- parib’ and ‘solid tumours’. The search was conducted by three independent reviewers (S.P.C., M.S. and O.B.), and a fourth reviewer was consulted in case of contro- versy (F.S.). No language restrictions were adopted. Some novel or updated results were published between August and December 2020 and were also included [51e53].

2.3. Data extraction

Details concerning study design, patient characteristics and current and previous treatment were extracted from each article, together with hazard ratios (HRs) and associated 95% confidence intervals (CIs) for PFS and OS, when reported, and the proportion of patients responding to evaluated treatments in each trial’s arms. These data had to be publicly available or computable from published articles/abstracts; otherwise, studies were excluded. Prespecified subgroup analyses for all end-points were performed, independent of the presence of heterogeneity, to highlight any differences between studies.
Some of the included studies provided only results for the intention-to-treat population, either BRCA-mut, BRCA-wt or with unknown BRCA status; others showed subgroup results for patients with BRCA- mut tumours and/or cancers with homologous recom- bination deficiency (HRD) due to other causes, as well. We collected and analysed the results for the overall population included in each study to avoid improper comparisons between overall and nested subpopulations in subgroup analyses.
We categorised the studies as per the following sub- groups: (1) BRCA1/2 mutational status (mut vs mixed/ wt), (2) tumour type (breast, ovarian, gastrointestinal, pancreatic, prostate cancer, nonesmall-cell lung cancer [NSCLC], small-cell lung cancer [SCLC] and mela- noma), (3) control protocol (CT placebo, placebo, enzalutamide or abiraterone, a PARPi without an antiangiogenic agent, an inferior dose of olaparib, bevacizumab), (4) treatment line (first-line treatment maintenance, second-line treatment, maintenance only), (5) different PARPi drugs (olaparib, talazoparib, veliparib, rucaparib, niraparib) and (6) tri- als’ phase (phase III vs II). Finally, an exploratory analysis on BRCA-independent HRD-positive tumours was carried out.

2.4. Data analysis

For the dichotomous variables (ORR), relative risks (RRs) with 95% CI were calculated for each study. The

time-to-event variables (PFS and OS) were analysed with HR and 95% CI. The Mantel-Haenszel method and the generic inverse-variance method were used to esti- mate RR and HR with their 95% CI, respectively. Heterogeneity among the studies was assessed using the c2-based Cochran Q statistic and the inconsistency index (I2 statistic) [54]. We preplanned to conduct the analyses using the random-effects (RE) model of Der- Simonian and Laird [55]. In case of non-significant heterogeneity, a fixed-effects (FE) model was subse- quently applied to confirm the result and perform the prespecified subgroup analyses [55]. To further investi- gate heterogeneity, we used the Baujat plot graphical method [56]. To assess the stability of the pooled results, multiple sensitivity analyses (influence analysis) were performed [57]. A more extensive explanation is re- ported in Supplementary methods.
Publication bias for each end-point was explored by
visual inspection of funnel plots, Egger’s regression test, Begg’s test and trim-and-fill analysis [58,59].
Data were analysed using R statistical software (version 4.0.2-packages: meta, metafor, dmetar) and RevMan 5.4 [60,61]. A two-tailed P-value 0.05 was considered statistically significant.
The risk of bias for each trial was assessed as per the criteria outlined in the Cochrane Handbook for Sys- tematic Reviews of Interventions [62]. Internal validity of eligible studies was assessed as per the Cochrane Collaboration’s Risk of Bias tool in Review Manager [61].

The study was registered in the Open Science Framework online repository ( (https://doi. org/10.17605/OSF.IO/NGY6D).

3. Results

Overall, 264 records were screened, and 29 studies (8,839 patients) met the inclusion criteria (Fig. 1).
Eight studies (27.6%) evaluated PARPis in the first- line setting, with 4 of them also including patients in more advanced lines of treatment (second-line treatment and/or further) and 1 studying a subsequent mainte- nance strategy. Thirteen studies (44.8%) included pa- tients in second-line treatment and/or further, and 8 (27.6%) more studies only focused on maintenance after first-line or more advanced lines of treatment. Fifteen (51.7%) studies were phase II RCTs, whereas the other 14 (48.3%) were phase III RCTs. Twelve (41.4%) studies included exclusively patients with BRCA1/2 mutation ; one (3.4%) study included a cohort with BRCA muta- tion and a cohort with BRCA-wt tumour. All other 16 (55.2%) studies recruited patients independent of their BRCA mutational status. Four (13.8%) studies recruited patients with breast cancer, 1 with (3.4%) colon cancer, 1 with (3.4%) gastric cancer, 2 (6.9%) with SCLC, 1 with (3.4%) NSCLC, 1 with (3.4%) melanoma, 4 (13.8%) with prostate cancer and 15 (51.9%) with ovarian cancer (some also including fallopian tube and primary peri- toneal cancers). The main study characteristics are re- ported in Table 1. Only 9 (31%) studies reported

Fig. 1. The PRISMA flow chart. ORR: overall response rate; PFS: progression-free survival; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses; OS: overall survival.

Table 1
Main characteristics of the included studies.
First author Year Journal Phase Line Cancer


No. of

No. of PTS,

No. of PTS, Treatments Meta-analysis





experimental arm

control arm


HRD dataa

Mirza 2016 New Engl J Med

III Maintenance after
≤second-line treatment

OC Mut 2 138 65 Niraparib vs placebo PFS Yes (PFS) Wt 2 234 116 Niraparib vs placebo PFS

Ledermann 2012/ 2016

New Engl J Med/Lancet Oncol

II Maintenance after
≤second-line treatment

OC Mixed 2 68 21 Olaparib vs placebo ORR, PFS, OS No


2019 New Engl J Med

III Maintenance after first- line treatment

OC Mixed 2 487 246 Niraparib vs placebo PFS, OS Yes (PFS)

Liu 2014/

Lancet Oncol/ II ≤Second-line treatment OC Mixed 2 46 44 Olaparib þ cediranib vs olaparib ORR, PFS, OS No
Ann Oncol

Oza 2015 Lancet Oncol II ≤Second-line treatment OC Mixed 2 81 81 Olaparib þ paclitaxel þ carboplatin– ORR, PFS, OS No
> olaparib vs paclitaxel þ carboplatin

Robson 2017/

New Engl J
Med/Ann Oncol

III ≤First-line treatment BC Mut 2 205 97 Olaparib vs chemotherapy ORR, PFS, OS No

Clarke 2018 Lancet Oncol II ≤Second-line treatment PC Mixed 2 71 71 Olaparib þ abiraterone vs
placebo þ abiraterone


Moore 2018 New Engl J Med
Golan 2019 New Engl J Med

III Maintenance after first- line treatment
III Maintenance after first- line treatment

OC Mut 2 260 131 Olaparib vs placebo PFS, OS No PC Mut 2 92 62 Olaparib vs placebo ORR, PFS, OS No

Ramalingam 2016 Clin Cancer

II First-line treatment NSCLC Mixed 2 105 53 Veliparib þ paclitaxel þ carboplatin
vs placebo þ paclitaxel þ carboplatin


Han 2018 Ann Oncol II ≤First-line treatment BC Mut 3 97 99 Veliparib þ carboplatin þ paclitaxel
vs paclitaxel þ carboplatin þ placebo
94 99 Veliparib þ temozolomide vs paclitaxel þ carboplatin þ placebo
Pietanza 2018 J Clin Oncol II ≤Second-line treatment SCLC Mixed 2 49 55 Temozolomide þ veliparib vs
temozolomide þ placebo


Coleman 2019 New Engl J

III First-line treatment and OC Mixed 3 382 375 Carboplatin þ paclitaxel þ veliparib– ORR, PFS Yes (PFS,


maintenance after first- line treatment

> veliparib vs
carboplatin þ paclitaxel–> placebo


Gorbunova 2019 Br J Cancer II First-line treatment CRC Mixed 2 65 65 FOLFIRI þ veliparib bevacizumab ORR, PFS, OS No
placebo þ FOLFIRI bevacizumab

Litton 2018 New Engl J Med

III ≤First-line treatment BC Mut 2 287 144 Talazoparib vs chemotherapy ORR, PFS, OS No

Owonikoko 2018 J Clin Oncol II First-line treatment SCLC Mixed 2 64 64 Cisplatin þ etoposide þ veliparib
cisplatin þ etoposide þ placebo


Middelton 2015 Ann Oncol II ≤Second-line treatment ME Mixed 3 116 115 Veliparib (20 mg) þ temozolomide vs ORR, PFS, OS No
temozolomide þ placebo
115 115 Veliparib (40 mg) þ temozolomide vs ORR, PFS, OS temozolomide þ placebo


liposomal doxorubicin

De Bono/

2020 New Engl J Med

III ≤Second-line treatment PC Mut 2 162 83 Olaparib vs enzalutamide/abiraterone ORR, PFS, OS Yes (PFS, OS)

Penson 2020 J Clin Oncol III ≤Third-line treatment OC Mut 2 178 88 Olaparib vs chemotherapy ORR, PFS No

Ray-Coquard 2019 New Engl J

III Maintenance after first- line treatment

OC Mixed 2 535 269 Olaparib þ bevacizumab vs
bevacizumab þ placebo


Mateo 2020 Lancet Oncol II ≤Second-line treatment PC Mut 2 49 49 Olaparib (400 mg twice) vs olaparib
(300 mg twice)
Audeh 2010 Lancet II ≤Second-line treatment OC Mut 2 33 24 Olaparib (400 mg twice) vs olaparib
(100 mg twice)



Dieras 2020 Lancet Oncol III First/second-line

BC Mut 2 337 172 Carboplatin þ paclitaxel þ veliparib
vs placebo þ carboplatin þ paclitaxel




Lancet Oncol/ III Maintenance after

OC Mut 2 196 99 Olaparib vs placebo PFS, OS No

Lauraine/ Poveda


J Clin Oncol

≤second-line treatment

OC: ovarian cancer; BC: breast cancer; PC: prostate cancer; NSCLC: nonesmall-cell lung cancer; SCLC: small-cell lung cancer; ME: melanoma; CRC: colorectal cancer; GC: gastric cancer; OS: overall survival; PFS: progression-free survival; ORR: overall response rate; Mut: mutant; Wt: wild-type; N: number; PTS: patients; FOLFIRI: 5-fluorouracil þ oxaliplatin þ irinotecan; HRD: homologous recombination deficiency.
a Homologous recombination deficiency not due to BRCA1/2 mutations.

subgroup analysis based on HRD status. Of these, 2 (22.3%) studies did not provide separate information for BRCA-independent and BRCA-dependent homologous recombination deficiency [30,45], 1 (11.1%) study re- ported data on ATM-negative gastric tumours [23], 3 (33.3%) studies reported separate data for a subpopu- lation of HRR-mut genes different from BRCA1/2 [38,39,46] and 3 (33.3%) studies reported separate data for a subgroup affected by BRCA-wt tumours, with HRD detected through the assessment of deleterious mutations in HRR genes, but also characteristic genomic scar signatures [29,31,44]. All of these studies reported PFS data, whereas only 3 reported OS [23,31,46] and ORR [23,29,30] results (Table 1).

3.1. Primary end-point: PFS

Overall, 26 studies provided data for PFS analysis, with 30 different comparisons. The pooled effect on PFS was statistically significant, with a considerable improve- ment provided by the experimental arms, although heterogeneity was high (HR: 0.59, 95% CI: 0.51e0.68, p < 0.001, I2 Z 85%) (Fig. 2).
No significant difference was observed with regard to
the patients’ mutational status (p Z 0.65), with a sig- nificant pooled effect observed in both mixed/wt cases (HR: 0.61, 95% CI: 0.52e0.71) and in patients with BRCA mutation (HR: 0.56, 95% CI: 0.42e0.75). Sig-
nificant subgroup differences were observed with regard to tumour site (p Z 0.001), line of therapy (p Z 0.002), control arm (p < 0.001), type of PARPi (p < 0.001) and

trials’ phase (p Z 0.006). More specifically, PFS was significantly improved in melanoma, SCLC, ovarian, prostate and pancreatic cancer, whereas the results were non-significant for NSCLC, breast and gastrointestinal cancers (Table 2 and Supplementary Fig. 1). PARPis improved PFS in all treatment lines, with a more pro- nounced effect observed in the maintenance setting, followed by second-line or further and first-line treat- ment (Table 2 and Supplementary Fig. 1). With respect to treatment comparison, PARPis were also superior to all control arms, with different degrees of benefit. The effect was more pronounced over placebo, followed by enzalutamide or abiraterone acetate (Enza/Abi) and CT placebo. Moreover, olaparib bevacizumab (beva) was superior to beva, and PARPi an anti- angiogenic drug (bevacizumab or cediranib) was supe- rior to a PARPi alone (Table 2 and Supplementary Fig. 1). All PARPis were effective, with a more pro- nounced benefit obtained with rucaparib, followed by niraparib, olaparib, talazoparib and veliparib (Table 2 and Supplementary Fig. 1).
Finally, a significant result was observed in both phase II and III trials, with a more pronounced benefit observed in the latter (Table 2 and Supplementary Fig. 1).
The main result, as well as numerous subgroup pooled estimates, was affected by significant heteroge- neity. As per the Baujat plot, the second comparison of the study from Han et al. [18] (i.e. ‘Han (2) 2018’), comparing veliparib þ temozolomide (VT) with carboplatin þ paclitaxel þ placebo (CPP) in breast

Fig. 2. Forest plot of progression-free survival. SE: standard error; HR: hazard ratio; IV: inverse-variance method; Random: random- effects model; CI: confidence interval.

Table 2
Progression-free survival results.
PFS No. of comparisons Pooled HR (95% CI) P pooled I2% P subgroups

Overall 30 0.59 (0.51e0.68) <0.001 85% N/A
Mutation status
Mixed/wild-type 17 0.61 (0.52e0.71) <0.001 79% 0.65
Mutant 13 0.56 (0.42e0.75) <0.001 89%
Tumour site
Ovarian 15 0.48 (0.40e0.58) <0.001 83% 0.001
Breast 5 0.77 (0.52e1.14) 0.19 88%
Prostate 2 0.46 (0.25e0.88) 0.02 85%
Melanoma 2 0.78 (0.62e0.98) 0.03 0%
NSCLC 1 0.72 (0.45e1.15) 0.17 N/A
SCLC 2 0.77 (0.63e0.95) 0.01 0%
Pancreatic 1 0.53 (0.35e0.80) 0.003 N/A
Gastrointestinal 2 0.86 (0.70e1.04) 0.12 0%
Line of therapy
First-line therapy maintenance 9 0.76 (0.61e0.93) 0.009 77% 0.002
≤Second-line therapy 13 0.61 (0.48e0.76) <0.001 83%
Maintenance only 8 0.42 (0.32e0.53) <0.001 83%
Control arm
CT placebo 17 0.75 (0.66e0.86) <0.001 63% <0.001
Placebo 8 0.39 (0.31e0.48) <0.001 77%
PARPi w/o an antiangiogenic drug 2 0.42 (0.27e0.64) <0.001 14%
Bevacizumab 1 0.63 (0.51e0.78) <0.001 N/A
Enzalutamide/abiraterone 2 0.46 (0.25e0.88) 0.02 85%
Type of PARPi
Olaparib 14 0.52 (0.42e0.64) <0.001 83% <0.001
Veliparib 10 0.82 (0.70e0.97) 0.02 63%
Niraparib 4 0.42 (0.29e0.60) <0.001 77%
Rucaparib 1 0.36 (0.30e0.43) <0.001 N/A
Talazoparib 1 0.54 (0.41e0.71) <0.001 N/A
Trials’ phase
Phase II 15 0.72 (0.59e0.88) 0.001 76% 0.006
Phase III 15 0.49 (0.41e0.59) <0.001 86%

PFS: progression-free survival; HR: hazard ratio; CI: confidence interval; N/A: not applicable; CT: chemotherapy; P pooled: p value of the pooled results; P subgroups: p values of the subgroup analyses; NSCLC: nonesmall-cell lung cancer; SCLC: small-cell lung cancer; PARPi: PARP in- hibitor; PARP: poly (ADP-ribose) polymerase.

cancer, and the study by Coleman et al. [29], comparing rucaparib with placebo in ovarian cancer, were the most relevant contributors to heterogeneity (Supplementary Fig. 2). A sensitivity analysis was performed to examine the stability and reliability of the pooled HR results. In the leave-one-out sensitivity analyses, the pooled overall effect estimate remained similar (data not shown). Considering the influence diagnostics plot (Supplementary Fig. 2), the study Han (2) 2018 was the most influential case. In fact, its omission from each subgroup improved the pooled effect estimates, as well as the main pooled PFS result (data not shown). Importantly, in the subgroup of breast cancer, the pooled HR showed a clinically meaningful and statisti- cally significant result, when omitting this study (HR:
0.62, 95% CI: 0.51e0.76, p < 0.001).

3.2. Secondary end-point: ORR

Overall, 24 studies provided data for ORR analysis, for a total of 27 comparisons. The pooled result showed a significant correlation between ORR and the

experimental arms (RR: 1.35, 95% CI: 1.16e1.56, p < 0.001, I2 Z 74%), with high heterogeneity (Fig. 3). The test for subgroup differences was non-significant with regard to tumour mutational status, with compa- rable effect observed for patients with BRCA mutation (RR: 1.44, 95% CI: 1.09e1.91) and the population of mixed or wt cases (RR: 1.27, 95% CI: 1.08e1.49). No difference was observed with regard to trials’ phase (p Z 0.09), with both phase II and III trials showing
significant association between the experimental arm and ORR (Table 3 and Supplementary Fig. 3). No sig- nificant difference was also observed with regard to tumour site (p Z 0.86), although the only study group with an individual statistically significant result was the one concerning ovarian cancer (p < 0.001).
Significant subgroup differences were observed with
regard to treatment line (p Z 0.03), control arm (p Z 0.04) and type of PARPi (p < 0.001). With respect to the first subgroup, a significant better association with ORR for the PARPi over the control was observed in second-line treatment and maintenance (Table 3 and Supplementary Fig. 3). PARPis showed a stronger association with ORR than with placebo and CT.

Fig. 3. Forest plot of overall response rates. SE: standard error; RR: relative risk; M-H: Mantel-Haenszel method; Random: random- effects model; CI: confidence interval.

Moreover, the combination with an antiangiogenic drug showed a significantly superior activity when compared with the same PARPi as a single agent. Conversely, there was no significant difference when PARPis were compared with Enza/Abi or when olaparib at a higher dose was compared with an inferior dose (Table 3 and Supplementary Fig. 3).
The effect was also significant with olaparib, nir- aparib and talazoparib but not with rucaparib or veli- parib (Table 3 and Supplementary Fig. 3).
In addition, in this case, the main result, as well as several subgroup pooled estimates, was affected by sig- nificant heterogeneity. As per the Baujat plot, the study by Litton et al. [19] (‘Litton 2018’) comparing talazo- parib with CT and, again, Han (2) 2018 [18] were the most relevant contributors to the observed heterogeneity (Supplementary Fig. 4). Both trials were focused on breast cancer.
In the leave-one-out sensitivity analyses, the pooled overall effect estimate remained similar also when removing the aforementioned studies (data not shown). Based on the influence diagnostics plot, Han (2) 2018 and Litton 2018 were considered potential influential cases (Supplementary Fig. 4). Although the main pooled effect remained significant when omitting the studies, subgroup results were affected within the breast cancer subset, wherein the omission of Han (2) 2018 led to a statistically significant result (RR: 1.54, 95% CI: 1.01e2.36, p Z 0.05). On the contrary, by removing Litton 2018, the result remained non-significant

(p Z 0.74). A similar influence was observed in the subgroup of first-line trials, wherein the removal of Han
(2) 2018 led to a statistically significant result (RR: 1.26, 95% CI: 1.03e1.53, p Z 0.02), whereas the removal of Litton 2018 did not impact the non-significance of the result (p Z 0.71). In the subgroup of control arms, when removing Litton 2018, the comparison with CT shifted to a non-significant result (p Z 0.08), whereas the removal of Han (2) 2018 strengthened the pooled result in favour of PARPi-based treatments, which however remained significant (RR: 1.27, 95% CI: 1.09e1.47, p Z 0.002). With respect to the PARPi subgroup, Litton 2018 was the only contributor to the talazoparib subset; however, when removing Han (2) 2018 from the veli- parib subgroup, the effect become only marginally non- significant (RR: 1.10, 95% CI: 1.00e1.22, p Z 0.06).

3.3. Secondary end-point: OS

Overall, 19 studies provided data for OS analysis, for a total of 22 comparisons. Pooled OS was significantly improved by the experimental arm (HR: 0.86, 95% CI: 0.80e0.93, p < 0.001, I2 Z 7%), with no significant heterogeneity (Fig. 4). Consequently, we performed again the analysis under the fixed-effects model, obtaining a comparable result (HR: 0.86, 95% CI: 0.80e0.92, p < 0.001, I2 Z 7%).
Because of the absence of substantial heterogeneity,
we performed prespecified subgroup analyses using the same FE model. No significant difference was observed

Table 3
Overall response rate results.
ORR No. of comparisons Pooled RR (95% CI) P pooled I2% P subgroups
Overall 27 1.35 (1.16e1.56) <0.001 74% N/A
Mutational status
1.27 (1.08e1.49)
Mutant 12 1.44 (1.09e1.91) 0.01 85%
Tumour site
1.42 (1.17e1.73)
Breast 5 1.24 (0.82e1.87) 0.30 93%
Prostate 3 1.75 (0.62e4.94) 0.29 78%
Melanoma 2 1.37 (0.73e2.54) 0.32 0%
NSCLC 1 0.93 (0.63e1.38) 0.97 N/A
SCLC 2 1.64 (0.59e4.53) 0.34 82%
Pancreatic 1 2.00 (0.85e4.70) 0.11 N/A
Gastrointestinal 1 1.17 (0.68e2.04) 0.57 78%
Line of therapy
First-line treatment maintenance
≤Second-line treatment Maintenance only
1.14 (0.90e1.41)
1.50 (1.23e1.82)
2.29 (1.28e4.07)
Control arm
CT placebo
1.20 (1.02e1.41)

Placebo 3 2.29 (1.28e4.07) 0.005 0%
Inferior dose of olaparib 2 1.55 (0.95e2.54) 0.08 11%
PARPi w/o an antiangiogenic drug 2 1.83 (1.37e2.45) <0.001 5%
Enzalutamide/abiraterone 2 3.17 (0.11e89.64) 0.50 90%
Type of PARPi
1.52 (1.26e1.84)
Veliparib 11 1.04 (0.89e1.22) 0.62 64%
Niraparib 1 2.28 (1.35e3.83) 0.002 N/A
Rucaparib 1 2.43 (0.98e6.06) 0.06 N/A
Talazoparib 1 2.30 (1.67e3.16) <0.001 N/A
Trials’ phase
Phase II
1.22 (1.01e1.47)
Phase III 9 1.63 (1.22e2.17) <0.001 87%
ORR: overall response rate; RR: relative risk; CI: confidence interval; CT: chemotherapy; P pooled: p value of the pooled results; P subgroups: p values of the subgroup analyses; NSCLC: nonesmall-cell lung cancer; SCLC: small-cell lung cancer; N/A: not applicable; PARPi: PARP inhibitor; PARP: poly (ADP-ribose) polymerase.

with regard to BRCA mutational status (p Z 0.57), tumour site (0.82), treatment line (p Z 0.22), control arm (p Z 0.21), PARPi (p Z 0.30) and trials’ phase (p Z 0.26).
Within subgroups, however, the subsets with an individually significant OS benefit associated with the experimental arms were both mixed/wt (p < 0.001) and BRCA-mut tumours (p Z 0.02), ovarian (p Z 0.004) and prostate cancer (p Z 0.04), second- line treatment (p Z 0.005) and maintenance (p Z 0.003), CT (p Z 0.02), Enza/Abi (p Z 0.04),
placebo (p Z 0.003), olaparib (p < 0.001) and both phase III (p < 0.001) and phase II RCT (p Z 0.05). Subgroup analysis results are detailed in Table 4 and
Supplementary Fig. 5.
When omitting each study in the leave-one-out sensitivity analysis, the overall result was never affected significantly (data not shown). The influence diagnostics plot identified Han (2) 2018 as a potential influential case (Supplementary Fig. 6). When reper- foming subgroup analyses by omitting it, the most affected subgroups were the one of tumour site,

treatment line, control arm, type of PARPi and RCT phase. More specifically, the pooled effect in breast cancer was improved and became only marginally non- significant (HR: 0.87, 95% CI: 0.75e1.01, p Z 0.07),
whereas the pooled effect in first-line treatment (HR: 0.87, 95% CI: 0.78e0.98, p Z 0.02), in the veliparib group (HR: 0.88, 95% CI: 0.78e1.00, p Z 0.05) and in the phase II RCT (HR: 0.85, 95% CI: 0.76e0.96,
p Z 0.006) became significant.

3.4. Subgroup analysis on HRD-positive tumours

We performed an exploratory subgroup analysis by pooling the treatment effects observed in the subpopu- lation of patients affected by HRD-positive tumours not exclusively owing to BRCA1/2 mutations. PARPi-based treatments appeared to be significantly effective in pro- longing PFS (HR: 0.51, 95% CI: 0.43e0.60, p < 0.001,
I2 Z 6%), with no significant heterogeneity observed
(Fig. 5). A numerical but non-significant correlation with higher response rates (RR: 1.57, 95% CI: 0.55e4.49, p Z 0.40, I2 Z 73%) and better OS (HR:



Fig. 4. Forest plot of overall survival. (A) Results under the random-effects model; (B) results under the fixed-effects model. SE: standard error; IV: inverse-variance method; Random: random-effects model; Fixed: fixed-effects model; CI: confidence interval.

0.85, 95% CI: 0.65e1.10, p Z 0.21, I2 Z 0%) compared
with the control arm was also observed, with significant heterogeneity for the former end-point and no hetero- geneity for the latter (Fig. 5).
Given the substantial absence of heterogeneity, we performed again the analyses under a fixed-effects model, obtaining comparable results in both PFS (HR: 0.51, 95% CI: 0.44e0.59, p < 0.001, I2 Z 6%) and OS
(HR: 0.85, 95% CI: 0.65e1.10, p Z 0.21, I2 Z 0%).

3.5. Risk-of-bias analysis and publication bias

The analysis of bias did not raise any specific concern. The only domain that showed higher risk, than the others, concerned the ‘performance bias’, which takes into account the blinding of study participants and personnel. In detail, 12 of 29 (41.4%) of the included studies were open-label ones. However, there were no, or very few, risk for other biases, suggesting an overall good internal validity of the studies included (Fig. 6 and Supplementary Fig. 7).
With respect to publication bias, the funnel plots of PFS and OS did not show asymmetry (Supplementary

Figs. 2 and 6), as also confirmed by non-significant Egger’s test (p Z 0.963 for PFS and p Z 0.599 for OS) and Begg’s test (p Z 0.832 for PFS and p Z 0.402 for OS). In the case of ORR, Egger’ test indicated the presence of funnel plot asymmetry (p Z 0.025), whereas Begg’s test was not significant (p Z 0.288). Therefore, we evaluated the effect of publication bias through a ‘trim-and-fill’ analysis (Supplementary Fig. 4). By using the L0 estimator, we obtained a significant pooled result (RR: 1.26, 95% CI: 1.10e1.46, p Z 0.001). A confir-
matory trim-and-fill analysis with another estimator (R0) showed similar results (RR: 1.33, 95% CI: 1.16e1.54, p < 0.001).

4. Discussion

4.1. Main results

Our study included 29 published phase II/III RCTs of metastatic solid tumours, wherein PARPi-containing regimens were compared with a therapeutic standard, represented in the majority of studies by either CT, hormonal treatment (HT) or placebo. Only a minority

Table 4
Overall survival results.
OS No. of comparisons Pooled HR (95% CI) P pooled I2% P subgroups
Overall RE 22 0.86 (0.80e0.93) <0.001 7% N/A
Overall FE 22 0.86 (0.80e0.92) <0.001 7% N/A
Mutation status
0.84 (0.76e0.93)
Mutant 11 0.88 (0.79e0.98) 0.02 23%
Tumour site
0.80 (0.69e0.93)
Breast 5 0.94 (0.82e1.08) 0.37 51%
Prostate 2 0.77 (0.59e0.99) 0.04 6%
Melanoma 2 0.89 (0.71e1.13) 0.35 6%
NSCLC 1 0.80 (0.54e1.19) 0.27 N/A
SCLC 1 0.83 (0.64e1.08) 0.16 N/A
Pancreatic 1 0.91 (0.56e1.48) 0.70 N/A
Gastrointestinal 2 0.85 (0.69e1.05) 012 60%
Line of therapy
First-line treatment maintenance
≤Second-line treatment Maintenance only
0.92 (0.82e1.03)
0.84 (0.74e0.95)
0.77 (0.66e0.91)
Control arm CT placebo Placebo
PARPi w/o an antiangiogenic drug
0.90 (0.83e0.99)
0.77 (0.66e0.91)
0.64 (0.36e1.14)
Type of PARPi
Olaparib 2

12 0.77 (0.59e0.99)

0.81 (0.73e0.90) 0.04

<0.001 6%


Veliparib 8 0.93 (0.83e1.05) 0.25 38%
Niraparib 1 0.70 (0.44e1.11) 0.13 N/A
Talazoparib 1 0.85 (0.67e1.07) 0.17 N/A
Trials’ phase
Phase II
0.90 (0.81e1.00)
Phase III 9 0.82 (0.74e0.91) <0.001 0%
RE: random-effects; FE: fixed-effects; OS: overall survival; HR: hazard ratio; CI: confidence interval; CT: chemotherapy; P pooled: p value of the pooled results; P subgroups: p values of the subgroup analyses; NSCLC: nonesmall-cell lung cancer; SCLC: small-cell lung cancer; PARPi: PARP inhibitor; PARP: poly (ADP-ribose) polymerase. Subgroup analyses were conducted under a FE model.

of studies (3/29) compared a PARPi in different doses (1 study) or compared the combination of a PARPi with an antiangiogenic drug vs the same PARPi alone (2 studies). Hence, our results substantially reflected the effect of PARPi-containing regimens against a different therapeutic standard.
The pooled analyses showed that PARPi regimens are associated with a consistent and statistically signifi- cant benefit in all clinical end-points, with an overall reduction in the instantaneous risk of progression of 41%, a strong association with ORR (RR: 1.35, p < 0.001) and ~14% reduction in the instantaneous risk of death. When observing prespecified subgroup analysis results, a differential PFS effect was observed with re-
gard to tumour site, line of therapy, the type of control arm, the type of PARPi and the trial phase. With respect to ORR, the treatment line, the type of control arm and PARPi were the subgroups showing statistically signifi- cant different within-subgroup results. Conversely, OS subgroup analyses did not identify subsets that might specifically benefit more than others. However, to note, significant individual subgroup results were observed for ovarian and prostate cancer, for olaparib (the most studied PARPi so far), in second or further lines of

treatment or maintenance, over CT, placebo and Enza/ Abi as the control, and in both phase II and III RCTs. To note, it is plausible that the lack of OS data in 10 of 29 studies limited the possibility to observe significant differences within subgroups.

4.2. Efficacy and activity based on the tumour type

When dissecting subgroup analyses, we observed that PARPi-based combinations seemed to be associated with prolonged PFS in ovarian cancer, prostate cancer, pancreatic cancer, melanoma and SCLC. In addition, after sensitivity analyses and the following selective removal of the VT vs CPP comparison [18], a clinically meaningful and statistically significant PFS improve- ment in the breast cancer subgroup was also observed, consistent with results from olaparib and talazoparib pivotal trials [19,21]. Apparently, the choice of temo- zolomide as the CT companion for veliparib in one BC trial turned out to produce such a profound significantly inferior performance for the experimental combination, such that the whole breast cancer subgroup pooled result was affected, despite being still numerically in favour of PARPis. Importantly, a significant association




Fig. 5. Pooled results of HRD-positive tumours. (A) PFS result; (B) ORR result; (C) OS result. OS: overall survival; PFS: progression-free survival; ORR: overall response rate; SE: standard error; IV: inverse-variance method; M-H: Mantel-Haenszel method; Random: random-effects model; CI: confidence interval; HRD: homologous recombination deficiency.

with better ORR was only observed for ovarian cancer, although the overall subgroup result did not show a statistical significance. At least two explanations might be given for this result. First, ovarian cancer regrouped the highest number of studies (11), whereas pooled re- sults for other cancer types relied only on 2 trials or 1 trial, with the exception of breast (5) and prostate (3) cancers. Second, ovarian cancer was proven to be particularly sensitive to PARPis because of a higher

prevalence of both BRCA-dependent and BRCA-inde- pendent HRD-positive tumours, than other solid tu- mours [63].

4.3. Efficacy and activity based on the treatment line and control arm

Experimental regimens improved PFS in all treatment lines, with a more profound effect in maintenance and

Fig. 6. Risk-of-bias analysis.

pretreated patients. However, the comparisons in first- line trials were mostly over CT (e.g. platinum-based regimens in ovarian cancer), whereas maintenance tri- als and some advanced line trials were against placebo, which might explain the differential effect observed. At the same time, the association with ORR was signifi- cant in advanced line and maintenance trials, but not in first-line trials. It is highly likely that this is the result of the higher concentration of CT control arms in the earlier-line RCT. This might appear contradic- tory with what was observed in the control arm sub- group, wherein PARPi regimens were superior to CT placebo. Despite this, it has to be considered that the most effective and active CT regimens are usually administered in upfront schedules. Hence, this might lead to differential effects observed on tumour shrinkage capabilities based on the treatment line. Similarly, the comparison with Enza/Abi in prostate cancer trials did not show a clear superiority for PARPis in terms of ORR. However, although several comparisons against CT involved a PARPi combined with a CT regimen, this was not the case for com- parisons against antiandrogen therapy (Table 1). This might potentially explain the reason of the comparable activity observed, but it could also mean that, because of the prominent growth driver role played by the androgen receptor pathway in prostate cancers [64], novel HT might retain their activity independent of the presence of HRD, despite PARPis being more effective in terms of PFS in this context. Importantly, compared with other solid tumours, patients with metastatic castration-resistant prostate cancer (mCRPC) present a higher prevalence of bone-only disease [65]. Therefore, some of the trials in this setting have used a composite end-point to evaluate response rates, by including also the percentage of prostate-specific antigen reduction from baseline and circulating tumour cell conversion (from more than/equal to 5 to less than 5) [47] or by including progression on the bone scan as per the Prostate Cancer Working Group criteria [45]. These differences might affect pooled ORR result interpre- tation, as well. In any case, when considering the ef- ficacy over different control arms, PARPi regimens were superior to all competing regimens in terms of PFS, with a more pronounced effect over placebo, over Enza/Abi (in prostate cancer) and over PARPi monotherapy when a combo with antiangiogenic drugs represented the experimental comparator (cediranib or bevacizumab).

4.4. Results based on mutational status

Notably, for all end-points (i.e. PFS, ORR and OS), no difference was observed between the subgroup of

patients with BRCA-mut and BRCA-wt/mixed tumours. This is a surprising, yet not completely unexpected finding. The first solid tumour wherein PARPis demonstrated a clear clinical benefit that translated into FDA approval was MOC, and half of the comparisons included in our analyses were conducted in this tumour type. Notably, individual pooled results for ovarian cancer were all uniformly in favour of PARPis in terms of PFS, ORR and OS, despite including numerous studies with BRCA-wt or unknown/mixed mutational status [29e31,34,36,38e40,44]. It is highly likely that such a benefit was driven by a subgroup of patients with HRD, a condition that can be caused by BRCA1/2 mutations, as well as by an impairment in other genes involved in the homologous recombination DNA repair mechanism, such as ATM, CHECK1/2, RAD51 or PALB2, either due to somatic/germinal mutations or due to epigenetic mechanisms [16,34,66]. Actually, roughly 50% of all high-grade serous ovarian cancers present some form of HRD because of germ line/so- matic mutations in BRCA1/2 (20%), epigenetic silencing of BRCA1 (11%), amplification/mutation of EMSY (8%), deletion of PTEN (7%), hypermethylation of RAD51C (3%) or mutations in ATM/ATR (2%) and Fanconi anaemia genes (5%) [67]. In this perspective, it is important to highlight that we performed an explor- atory subgroup analysis on HRD-positive tumours, which comprised 9 studies. Of these, 6 were on MOC [29e31,38,39,44], 2 were on mCRPC [45,46] and 1 was on gastric cancer [23]. We observed a strikingly 49% significant reduction in the risk of progression or death with PARPi-based treatments, compared with the con- trol. This result was undoubtedly driven by MOC studies, but also 1 of 2 mCRPC studies showed a sig- nificant result in favour of PARPis. This result strengthens the arising theory that PARPis might be particularly effective not only in BRCA-mut tumours but also in tumours with other forms of defective HRR. Unfortunately, to date, the definition of HRD is extremely heterogeneous, and several clinical trials have used different methods to assess it [47,68,69]. Therefore, the implementation of a more homogeneous characterisation of HRD status across different solid tumours is highly recommended.
Intriguingly, the Preferred Reporting Items for Sys-
tematic Reviews and Meta-Analyses (PRISMA) trial in ovarian cancer was able to identify a PFS improvement with niraparib monotherapy in HRR-proficient MOC [31]. Similarly, a study of olaparib abiraterone vs abiraterone in mCRPC showed improved radiologic PFS irrespective of HRR status, with exploratory ana- lyses suggesting efficacy also in non-dysfunctional tu- mours [45]. In addition, the combination of veliparib, cisplatin and etoposide in the ECO-ACRIN 2511 study

in SCLC also showed a significant PFS improvement in the absence of HRD, possibly owing to a synergistic effect with CT agents capable of directly damaging DNA [24,30,41]. Nevertheless, some individual trials still failed to demonstrate the efficacy of PARPis in non- mut tumours such as melanoma [27], NSCLC and SCLC [25,26], and colon and gastric cancer [22,23], albeit PARPis had been combined with several effective CT partners. Given the low frequency of HRR genes such as BRCA1/2 in tumours such as melanoma, colon cancer, gastric cancer, SCLC and NSCLC [7,66], it is not particularly surprising that some results observed in unselected populations have been disappointing. Still, there is preclinical evidence for potential alternative biomarkers of response to PARPis in subgroups of these solid tumours, such as low ERCC1 expression in NSCLC and melanoma [27,70], ARID1A deficiency in solid tumours, including gastric and colon cancer [71], detectable p16 expression in melanoma [27] or bio- markers of resistance, such as TRIP12, which has been recently demonstrated to constrain the PARP1 trapping mechanism of PARPis [72]. A more extensive evaluation of these biomarkers in future studies, so as to better select the target population for PARPis in BRCA-wt solid tumours, is highly recommended. Furthermore, recent findings have shown that the overall frequency of mutations affecting HRR genes is around 17%, with the maximum prevalence observed in endometrial cancer (34.4%) and the lowest observed in gastrointestinal stromal tumours (3.7%) [66].
In light of our results supporting PARPi efficacy also in BRCA-independent HRD-positive solid tumours, a potential way to assess the efficacy of PARPis in rarer BRCA-mut or BRCA-wt/HRD-positive cancers might be the development of basket trials, with the objective to prove a class effect as a tumour-agnostic therapeutic option. This has been already observed, for example, with NTRK fusionepositive or with high microsatellite instability tumours [73,74]. Few trials of this kind are already planned/ongoing (i.e. NCT03742895, NCT04123366, NCT04171700, NCT04503265, NCT04174716).
Another possibility to overcome patient recruitment
issues in trials involving rare tumour types might be through the comparison of single-arm trials involving the experimental drug with a synthetic control arm represented, for example, by historical observational data, already published results from RCTs or other external control data [75]. This strategy is not new but is gaining more attention in recent years, having led, among others, to the expanded indication of palbociclib HT for men with hormone receptore- positive/human epidermal growth factor receptor 2 (HER2)-negative MBC [75].

4.5. Results based on the PARPi molecule

All PARPis prolonged PFS. The most pronounced ef- fect was observed with rucaparib, whereas the less potent effect was observed with veliparib. Similarly, for ORR, the most potent PARPi was rucaparib and the less was veliparib. However, individually, rucaparib and veliparib pooled results were non-significant. It is important to underline that the PARPi’s main thera- peutic effect seems to be related to PARylation and subsequent PARP trapping [13,16]. In this perspective, different PARPis have shown different trapping potency, with talazoparib>niraparib>olaparib Z
rucaparib>veliparib, the latter substantially lacking
PARP trapping capability [16,76,77]. This is well rep-
resented by the poor performance observed in our pooled analyses, despite the numerous veliparib- containing studies included (11 of 29). In apparent contrast, rucaparib has provided the best individual result in terms of PFS and ORR (OS data were un- available), compared with the other PARPis. However, it is highly likely that this result is attributable to the fact that only one trial contained rucaparib, and in this study, the PARPi was compared with placebo as maintenance treatment after response to platinum agents in ovarian cancer [29]. Good responses to plat- inum agents, at least in MOC and mCRPC, have been linked to the presence of HRD [78,79] and the potential to be a predictor of response to PARP inhibition [78]; thus, a particularly sensitive population might have been tested. At the same time, no active comparison was administered. As a consequence, different from ruca- parib, the effect of other PARPis might have been diluted in pooled analyses, regrouping several studies with different combinations, comparisons and tumour types. This is not true for talazoparib, the most potent PARPi in terms of PARP trapping, which was also tested in only 1 trial included in our study and still did not outperform the other inhibitors. However, it is necessary to underline that it was administered in a poor-prognosis breast cancer subgroup (i.e. triple negative), and different from rucaparib, it was compared with potentially effective therapeutic alternatives, such as eribulin and capecitabine [19]. In any case, our results substantially confirm the superiority of PARPis with trapping capacity over veliparib.
In addition, some other mechanisms of action for PARPis have been proposed, such as the blocking of PARP-regulated gene transcription, interference with ribosome biogenesis, mitophagy and apoptosis, which might differ between different PARPi molecules and might show different impact in different types of cancers [38,80e83]. In addition to this, other molecular and cellular mechanisms (e.g. immune pathway activation,

programmed death-ligand 1 (PD-L1) expression modu- lation on cancer cells, genomic instability produced by PARP inhibition) might increase tumour immunoge- nicity and responsiveness to immune checkpoint in- hibitors (ICIs) [84e86]. In this perspective, promising evidence of efficacy for the PARPi ICI combination has been recently observed with the TOPACIO and MEDIOLA single-arm phase II trials, and RCTs are already ongoing [7,49,87,88].
All in all, a deeper characterisation of all these mechanisms is warranted, so as to identify the best combination strategies and the most adequate PARPi for the appropriate context.

4.6. Limitations and strengths

The major limitation of our study relies on the considerable heterogeneity observed for PFS and ORR pooled results. It is highly likely that such a hetero- geneity was related to the design of the study itself, having included in our analysis trials of different phases, conducted in different lines, conducted on several solid tumours, on both mut and wt populations and with different PARPis and control arms. In fact, subgroup analyses identified specific subsets wherein the efficacy and activity of PARPi-based regimens seem to be modest with respect to the therapeutic standard. Importantly, when performing leave-one-out sensitivity analyses, the main pooled results were not affected significantly by a single study/treatment com- parison. Moreover, an RE model was applied to take into account such heterogeneity. We addressed it also through visual inspection of Baujat plots and influen- tial analyses, which helped to identify the most prob- lematic comparisons and assess their impact on each subgroup.
Another limitation is represented by the use of in- dividual patient data (IPD). We were not able to perform an IPD meta-analysis because of the lack of the necessary resources. Although this kind of meta- analysis is usually considered the best option to sum- marise the results of multiple studies, the scientific literature recognises that such studies are not always feasible [89]. Moreover, although some guidance is available to help understand when aggregate patient data (APD) meta-analyses, such as ours, might suffice and when IPD might add value, this is not backed up by empirical evidence [90], and it is still not clear when the collection of more detailed IPD is truly needed [91]. In fact, for meta-analyses of published time-to-event outcomes, individual case studies have shown that they can produce effects that are either larger than, smaller than or similar to their IPD equivalents [90]. Moreover, HRs from published APD meta-analyses seem to most likely agree with those from IPD when the information size is large [91]. Finally, considering the complexity of the topic and the high number of

studies and patients involved, it is highly likely that an IPD meta-analysis on the same topic will not be conducted.
Another limitation is represented by the publication bias observed regarding the ORR result. Importantly, however, based on the trim-and-fill analyses performed, the ORR pooled result was confirmed to be statistically significant even when controlling for selective publica- tion, thus suggesting that this bias had little effect and the results were relatively robust. Finally, we did not analyse here the toxicity data emerging from those trials. In any case, PARPis are usually well-tolerated drugs, with nausea, vomiting, seizures, fatigue, leukopenia, anaemia and thrombocytopenia being the most frequent, albeit manageable side-effects. The incidence is different with respect to the PARPi molecule, as also well described elsewhere [7,77,92].
The strength of our study relies on its comprehensive assessment of PARPi activity and efficacy in solid tu- mours, which is the most complete and up-to-date one. The methodology was solid and reliable, with numerous sensitivity analyses conducted to overcome the main issues observed related to heterogeneity and robustness of results.

5. Conclusions

Although our study confirms and reinforces the role of already approved PARPi-based treatments, especially in BRCA1/2-mut tumours, a more comprehensive effort is needed to identify other forms of HRD along with a better characterisation of secondary mechanisms of ac- tion and further predictive biomarkers of response. We envision that this approach will better elucidate PARPi efficacy in a broader scenario, alone or in combination with other therapeutic agents.

Data availability statement

The database for the analyses of this study is available on request from the corresponding author, although, in general, the results are publicly available and retrievable from the included studies.

Code availability statement

The R codes used for the analyses are available from the corresponding author on reasonable request.

Author contributions

F.S. and D.G. conceived the study. S.P.C., O.B., M.S. and F.S. performed the systematic review of the litera- ture. F.G. and F.S. performed the statistical analyses. All authors contributed to interpreting the data. F.S., F.G., D.G. and P.R. wrote the first manuscript draft,

and all authors contributed to writing, correcting and approving the final version of the manuscript.

Ethical prescription for data collection and management

Not applicable.


P.R. is supported by a Prostate Cancer Foundation’s Young Investigator award. F.S. is supported by a Eu- ropean Society for Medical Oncology (ESMO) Trans- lational Research Fellowship.

Conflict of interest statement

M.G. and S.D.P. have declared honoraria from Roche, Pfizer, AstraZeneca, Novartis, Celgene, Eli Lilly, Amgen and Eisai. A.P. has declared an immediate family member being employed by Novartis; personal honoraria from Pfizer, Novartis, Roche, MSD Oncology, Lilly and Daiichi Sankyo; travel, accommo- dations and expenses paid by Daiichi Sankyo; research funding from Roche and Novartis; a consulting/advi- sory role for NanoString Technologies, Amgen, Roche, Novartis, Pfizer and Bristol Myers Squibb and patent PCT/EP2016/080056: HER2 AS A PREDICTOR OF RESPONSE TO DUAL HER2 BLOCKADE IN THE ABSENCE OF CYTOTOXIC THERAPY. G.C. is an expert testimony for Pfizer, Novartis and Roche Gen- entech and a member of the steering committee for randomised clinical trials of Cascadian, Roche Gen- entech and MacroGenics. D.G. has declared consulting fees from Novartis, Lilly and Pfizer and research fund- ing from LILT, Novartis, AstraZeneca and the Uni- versity of Trieste. I.P. has declared consulting fees from Roche, Novartis, Lilly, Pfizer, AstraZeneca, Pierre Fabre and Ipsen. G.S. has declared grant/research Support from MSD Italia S.r.l. and a consulting role for TESARO Bio Italy S.r.l., Johnson & Johnson and Clovis Oncology Italy S.r.l. The other authors have nothing to declare.


The authors are thankful to Alejandra Mej´ıa Arango for language editing.

Appendix A. Supplementary data

Supplementary data to this article can be found online at


[1] National Cancer Institute. Genetics of breast and gynecologic cancers (PDQ®) – health professional version. 2019.
[2] Venkitaraman AR. Cancer susceptibility and the functions of BRCA1 and BRCA2. Cell 2002;108(2):171e82.
[3] Featherstone C, Jackson SP. DNA double-strand break repair. Curr Biol 1999;9(20):R759e61.
[4] Wang JYJ. DNA damage and apoptosis. Cell Death Differ 2001; 8(11):1047e8.
[5] Narod SA, Foulkes WD. BRCA1 and BRCA2: 1994 and beyond. Nat Rev Cancer 2004;4(9):665e76.
[6] Susan G. Komen foundation. Inherited genetic mutations.
[7] Bayraktar S, Glu¨ ck S, Darling H. Update on PARP inhibitor therapy for solid tumours. J Cancer Prev Curr Res 2019;10(4): 98e107.
[8] Golan T, Hammel P, Reni M, et al. Maintenance olaparib for germline BRCA-mutated metastatic pancreatic cancer. N Engl J Med 2019;381(4):317e27.
[9] McClain MR, Palomaki GE, Nathanson KL, Haddow JE. Adjusting the estimated proportion of breast cancer cases asso- ciated with BRCA1 and BRCA2 mutations: public health impli- cations. Genet Med Off J Am Coll Med Genet 2005;7(1):28e33.
[10] Messina C, Cattrini C, Soldato D, et al. BRCA mutations in prostate cancer: prognostic and predictive implications. J Oncol 2020.
[11] Manchana T, Phoolcharoen N, Tantbirojn P. BRCA mutation in high grade epithelial ovarian cancers. Gynecol Oncol Rep 2019; 29:102e5.
[12] Satoh MS, Lindahl T. Role of poly(ADP-ribose) formation in DNA repair. Nature 1992;356(6367):356e8.
[13] Lord CJ, Ashworth A. PARP inhibitors: synthetic lethality in the clinic. Science 2017;355(6330):1152e8.
[14] Farmer H, McCabe N, Lord CJ, et al. Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy. Nature 2005;434(7035):917e21.
[15] McCabe N, Turner NC, Lord CJ, et al. Deficiency in the repair of DNA damage by homologous recombination and sensitivity to poly(ADP-ribose) polymerase inhibition. Cancer Res 2006;66(16): 8109e15.
[16] Murai J, Huang SN, Das BB, et al. Trapping of PARP1 and PARP2 by clinical PARP inhibitors. Cancer Res 2012;72(21): 5588e99.
[17] Hopkins TA, Ainsworth WB, Ellis PA, et al. PARP1 trapping by PARP inhibitors drives cytotoxicity in both cancer cells and healthy bone marrow. Mol Cancer Res 2019;17(2):409e19.
[18] Han HS, Die´ras V, Robson M, et al. Veliparib with temozolomide or carboplatin/paclitaxel versus placebo with carboplatin/pacli- taxel in patients with BRCA1/2 locally recurrent/metastatic breast cancer: randomized phase II study. Ann Oncol Off J Eur Soc Med Oncol 2018;29(1):154e61.
[19] Litton JK, Rugo HS, Ettl J, et al. Talazoparib in patients with advanced breast cancer and a germline BRCA mutation. N Engl J Med 2018;379(8):753e63.
[20] Robson ME, Tung N, Conte P, et al. OlympiAD final overall survival and tolerability results: olaparib versus chemotherapy treatment of physician’s choice in patients with a germline BRCA mutation and HER2-negative metastatic breast cancer. Ann Oncol Off J Eur Soc Med Oncol 2019;30(4):558e66.
[21] Robson M, Im S-A, Senkus E, et al. Olaparib for metastatic breast cancer in patients with a germline BRCA mutation. N Engl J Med 2017;377(6):523e33.
[22] Gorbunova V, Beck JT, Hofheinz R-D, et al. A phase 2 rando- mised study of veliparib plus FOLFIRI bevacizumab versus

placebo plus FOLFIRI bevacizumab in metastatic colorectal cancer. Br J Cancer 2019;120(2):183e9.
[23] Bang Y-J, Xu R-H, Chin K, et al. Olaparib in combination with paclitaxel in patients with advanced gastric cancer who have progressed following first-line therapy (GOLD): a double-blind, randomised, placebo-controlled, phase 3 trial. Lancet Oncol 2017;18(12):1637e51.
[24] Owonikoko TK, Dahlberg SE, Sica GL, et al. Randomized phase II trial of cisplatin and etoposide in combination with veliparib or placebo for extensive-stage small-cell lung cancer: ECOG-ACRIN 2511 study. J Clin Oncol Off J Am Soc Clin Oncol 2019;37(3):222e9.
[25] Pietanza MC, Waqar SN, Krug LM, et al. Randomized, double- blind, phase II study of temozolomide in combination with either veliparib or placebo in patients with relapsed-sensitive or re- fractory small-cell lung cancer. J Clin Oncol Off J Am Soc Clin Oncol 2018;36(23):2386e94.
[26] Ramalingam SS, Blais N, Mazieres J, et al. Randomized, placebo- controlled, phase II study of veliparib in combination with car- boplatin and paclitaxel for advanced/metastatic non-small cell lung cancer. Clin Cancer Res Off J Am Assoc Cancer Res 2017; 23(8):1937e44.
[27] Middleton MR, Friedlander P, Hamid O, et al. Randomized phase II study evaluating veliparib (ABT-888) with temozolomide in patients with metastatic melanoma. Ann Oncol Off J Eur Soc Med Oncol 2015;26(10):2173e9.
[28] Audeh MW, Carmichael J, Penson RT, et al. Oral poly(ADP- ribose) polymerase inhibitor olaparib in patients with BRCA1 or BRCA2 mutations and recurrent ovarian cancer: a proof-of- concept trial. Lancet Lond Engl 2010;376(9737):245e51.
[29] Coleman RL, Oza AM, Lorusso D, et al. Rucaparib maintenance treatment for recurrent ovarian carcinoma after response to platinum therapy (ARIEL3): a randomised, double-blind, pla- cebo-controlled, phase 3 trial. Lancet Lond Engl 2017;390(10106): 1949e61.
[30] Coleman RL, Fleming GF, Brady MF, et al. Veliparib with first- line chemotherapy and as maintenance therapy in ovarian cancer. N Engl J Med 2019;381(25):2403e15.
[31] Gonza´lez-Mart´ın A, Pothuri B, Vergote I, et al. Niraparib in patients with newly diagnosed advanced ovarian cancer. N Engl J Med 2019;381(25):2391e402.
[32] Kaye SB, Lubinski J, Matulonis U, et al. Phase II, open-label, randomized, multicenter study comparing the efficacy and safety of olaparib, a poly (ADP-ribose) polymerase inhibitor, and pegylated liposomal doxorubicin in patients with BRCA1 or BRCA2 mutations and recurrent ovarian cancer. J Clin Oncol Off J Am Soc Clin Oncol 2012;30(4):372e9.
[33] Kummar S, Oza AM, Fleming GF, et al. Randomized trial of oral cyclophosphamide and veliparib in high-grade serous ovarian, primary peritoneal, or fallopian tube cancers, or BRCA-mutant ovarian cancer. Clin Cancer Res Off J Am Assoc Cancer Res 2015;21(7):1574e82.
[34] Ledermann J, Harter P, Gourley C, et al. Olaparib maintenance therapy in platinum-sensitive relapsed ovarian cancer. N Engl J Med 2012;366(15):1382e92.
[35] Ledermann JA, Harter P, Gourley C, et al. Overall survival in patients with platinum-sensitive recurrent serous ovarian cancer receiving olaparib maintenance monotherapy: an updated anal- ysis from a randomised, placebo-controlled, double-blind, phase 2 trial. Lancet Oncol 2016;17(11):1579e89.
[36] Liu JF, Barry WT, Birrer M, et al. Combination cediranib and olaparib versus olaparib alone for women with recurrent platinum-sensitive ovarian cancer: a randomised phase 2 study. Lancet Oncol 2014;15(11):1207e14.
[37] Liu JF, Barry WT, Birrer M, et al. Overall survival and updated progression-free survival outcomes in a randomized phase II study of combination cediranib and olaparib versus olaparib in relapsed platinum-sensitive ovarian cancer. Ann Oncol Off J Eur Soc Med Oncol 2019;30(4):551e7.
Mirza MR, Monk BJ, Herrstedt J, et al. Niraparib maintenance therapy in platinum-sensitive, recurrent ovarian cancer. N Engl J Med 2016;375(22):2154e64.
[39] Mirza MR, A˚ vall Lundqvist E, Birrer MJ, et al. Niraparib plus
bevacizumab versus niraparib alone for platinum-sensitive recurrent ovarian cancer (NSGO-AVANOVA2/ENGOT-ov24): a rando- mised, phase 2, superiority trial. Lancet Oncol 2019;20(10):1409e19.
[40] Moore K, Colombo N, Scambia G, et al. Maintenance olaparib in patients with newly diagnosed advanced ovarian cancer. N Engl J Med 2018;379(26):2495e505.
[41] Oza AM, Cibula D, Benzaquen AO, et al. Olaparib combined with chemotherapy for recurrent platinum-sensitive ovarian can- cer: a randomised phase 2 trial. Lancet Oncol 2015;16(1):87e97.
[42] Penson RT, Valencia RV, Cibula D, et al. Olaparib versus non- platinum chemotherapy in patients with platinum-sensitive relapsed ovarian cancer and a germline BRCA1/2 mutation (SOLO3): a randomized phase III trial. J Clin Oncol Off J Am Soc Clin Oncol 2020;38(11):1164e74.
[43] Pujade-Lauraine E, Ledermann JA, Selle F, et al. Olaparib tablets as maintenance therapy in patients with platinum-sensitive, relapsed ovarian cancer and a BRCA1/2 mutation (SOLO2/- ENGOT-Ov21): a double-blind, randomised, placebo-controlled, phase 3 trial. Lancet Oncol 2017;18(9):1274e84.
[44] Ray-Coquard I, Pautier P, Pignata S, et al. Olaparib plus bev- acizumab as first-line maintenance in ovarian cancer. N Engl J Med 2019;381(25):2416e28.
[45] Clarke N, Wiechno P, Alekseev B, et al. Olaparib combined with abiraterone in patients with metastatic castration-resistant pros- tate cancer: a randomised, double-blind, placebo-controlled, phase 2 trial. Lancet Oncol 2018;19(7):975e86.
[46] de Bono J, Mateo J, Fizazi K, et al. Olaparib for metastatic castration-resistant prostate cancer. N Engl J Med 2020;382(22): 2091e102.
[47] Mateo J, Porta N, Bianchini D, et al. Olaparib in patients with metastatic castration-resistant prostate cancer with DNA repair gene aberrations (TOPARP-B): a multicentre, open-label, rand- omised, phase 2 trial. Lancet Oncol 2020;21(1):162e74.
[48] Poveda A, Floquet A, Ledermann JA, et al. Final overall survival (OS) results from SOLO2/ENGOT-ov21: a phase III trial assessing maintenance olaparib in patients (pts) with platinum- sensitive, relapsed ovarian cancer and a BRCA mutation. J Clin Oncol 2020;38(15_suppl). 6002e6002.
[49] Vikas P, Borcherding N, Chennamadhavuni A, Garje R. Thera- peutic potential of combining PARP inhibitor and immuno- therapy in solid tumours. Front Oncol 2020. https:
[50] Clinical trial endpoints for the approval of cancer drugs and bi- ologics. 2018. p. 19.
[51] Hussain M, Mateo J, Fizazi K, et al. Survival with olaparib in metastatic castration-resistant prostate cancer. N Engl J Med 2020.
[52] Die´ras V, Han HS, Kaufman B, et al. Veliparib with carboplatin and paclitaxel in BRCA-mutated advanced breast cancer (BROCADE3): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol 2020;21(10):1269e82.
[53] Litton JK, Hurvitz SA, Mina LA, et al. Talazoparib versus chemotherapy in patients with germline BRCA1/2-mutated HER2-negative advanced breast cancer: final overall survival re- sults from the EMBRACA trial. Ann Oncol Off J Eur Soc Med Oncol 2020;31(11):1526e35.
[54] Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21(11):1539e58.
[55] Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. A basic introduction to fixed-effect and random-effects models for meta- analysis. Res Synth Methods 2010;1(2):97e111.
[56] Baujat B, Mahe´ C, Pignon J-P, Hill C. A graphical method for exploring heterogeneity in meta-analyses: application to a meta- analysis of 65 trials. Stat Med 2002;21(18):2641e52.

[57] Viechtbauer W, Cheung MW-L. Outlier and influence diagnostics for meta-analysis. Res Synth Methods 2010;1(2):112e25.
[58] Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommenda- tions for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011;343: d4002.
[59] Sterne JA, Egger M. Funnel plots for detecting bias in meta- analysis: guidelines on choice of axis. J Clin Epidemiol 2001; 54(10):1046e55.
[60] R Core Team. R. A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2017.
[61] The Cochrane Collaboration. Review manager (RevMan). Copenhagen: The Nordic Cochrane Centre; 2014.
[62] Higgins J, Green S. Cochrane Handbook for systematic reviews of Interventions Version 5.1.0 [updated March 2011], The Cochrane Collaboration.
[63] Nguyen L, Martens JWM, Van Hoeck A, Cuppen E. Pan-cancer landscape of homologous recombination deficiency. Nat Com- mun 2020;11(5584).
[64] Lonergan PE, Tindall DJ. Androgen receptor signaling in pros- tate cancer development and progression. J Carcinog 2011. https:
[65] Hernandez RK, Wade SW, Reich A, et al. Incidence of bone metastases in patients with solid tumours: analysis of oncology electronic medical records in the United States. BMC Cancer 2018.
[66] Heeke AL, Pishvaian MJ, Lynce F, et al. Prevalence of homolo- gous recombinationerelated gene mutations across multiple cancer types. JCO Precis Oncol 2018. https:
[67] Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature 2011;474(7353):609e15.
[68] Myriad Genetics, Inc. Myriad myChoice® CDx description. Available at: medicine/mychoice-cdx/. 2020.
[69] Foundation Medicine, Inc. FoundationOne®CDx technical specifications. 2018. Available at: www.rochefoundationmedicine. com/f1cdxtech. 2020.
[70] Postel-Vinay S, Bajrami I, Friboulet L, et al. A high-throughput screen identifies PARP1/2 inhibitors as a potential therapy for ERCC1-deficient non-small cell lung cancer. Oncogene 2013; 32(47):5377e87.
[71] Shen J, Peng Y, Wei L, et al. ARID1A deficiency impairs the DNA damage checkpoint and sensitizes cells to PARP inhibitors. Cancer Discov 2015;5(7):752e67.
[72] Gatti M, Imhof R, Huang Q, et al. The ubiquitin ligase TRIP12 limits PARP1 trapping and constrains PARP inhibitor efficiency. Cell Rep 2020.
[73] FDA Approves Entrectinib for Tumours with NTRK Fusions – National Cancer Institute. [ cancer-currents-blog/2019/fda-entrectinib-ntrk-fusion]. .
[74] FDA approves pembrolizumab for first-line treatment of MSI- H/dMMR colorectal cancer. [ approvals-and-databases/fda-approves-pembrolizumab-first-line- treatment-msi-hdmmr-colorectal-cancer]. .
[75] Thorlund K, Dron L, Park JJH, Mills EJ. Synthetic and external controls in clinical trials e a primer for researchers. Clin Epi- demiol 2020;12:457e67.
[76] Murai J, Huang S-YN, Renaud A, et al. Stereospecific PARP trapping by BMN 673 and comparison with olaparib and ruca- parib. Mol Cancer Ther 2014;13(2):433e43.
Pommier Y, O’Connor MJ, de Bono J. Laying a trap to kill cancer cells: PARP inhibitors and their mechanisms of action. Sci Transl Med 2016;8(362):362ps17.
[78] Jiang X, Li X, Li W, et al. PARP inhibitors in ovarian cancer: sensitivity prediction and resistance mechanisms. J Cell Mol Med 2019;23(4):2303e13.
[79] Zafeiriou Z, Bianchini D, Chandler R, et al. Genomic analysis of three metastatic prostate cancer patients with exceptional re- sponses to carboplatin indicating different types of DNA repair deficiency. Eur Urol 2019;75(1):184e92.
[80] Kim D-S, Camacho CV, Nagari A, et al. Activation of PARP-1 by snoRNAs controls ribosome biogenesis and cell growth via the RNA helicase DDX21. Mol Cell 2019;75(6): 1270e1285.e14.
[81] Brenner JC, Ateeq B, Li Y, et al. Mechanistic rationale for inhi- bition of poly(ADP-ribose) polymerase in ETS gene fusion- positive prostate cancer. Cancer Cell 2011;19(5):664e78.
[82] Arun B, Akar U, Gutierrez-Barrera AM, et al. The PARP in- hibitor AZD2281 (Olaparib) induces autophagy/mitophagy in BRCA1 and BRCA2 mutant breast cancer cells. Int J Oncol 2015; 47(1):262e8.
[83] Lee EK, Konstantinopoulos PA. PARP inhibition and immune modulation: scientific rationale and perspectives for the treatment of gynecologic cancers. Ther Adv Med Oncol 2020. https:
[84] Chabanon RM, Soria J-C, Lord CJ, Postel-Vinay S. Beyond DNA repair: the novel immunological potential of PARP in- hibitors. Mol Cell Oncol 2019. https:
[85] Hodgson DR, Dougherty BA, Lai Z, et al. Candidate biomarkers of PARP inhibitor sensitivity in ovarian cancer beyond the BRCA genes. Br J Cancer 2018;119(11):1401e9.
[86] Gay CM, Byers LA. PARP inhibition combined with immune checkpoint blockade in SCLC: oasis in an immune desert or mirage? J Thorac Oncol Off Publ Int Assoc Study Lung Cancer 2019;14(8):1323e6.
[87] Domchek SM, Postel-Vinay S, Im S-A, et al. Olaparib and dur- valumab in patients with germline BRCA-mutated metastatic breast cancer (MEDIOLA): an open-label, multicentre, phase 1/2, basket study. Lancet Oncol 2020;21(9):1155e64.
[88] Konstantinopoulos PA, Waggoner SE, Vidal GA, et al. TOPA- CIO/Keynote-162 (NCT02657889): a phase 1/2 study of niraparib þ pembrolizumab in patients (pts) with advanced triple-negative breast cancer or recurrent ovarian cancer (ROC)dresults from ROC cohort. J Clin Oncol 2018;36(15_suppl). 106e106.
[89] Berlin JA, Santanna J, Schmid CH, et al. Individual patient- versus group-level data meta-regressions for the investigation of treatment effect modifiers: ecological bias rears its ugly head. Stat Med 2002;21(3):371e87.
[90] Tudur Smith C, Clarke M, Marson T, et al. A framework for deciding if individual participant data are likely to be worthwhile. Abstracts of the 23rd Cochrane Colloquium, Vienna, Austria. Cochrane Database Syst Rev 2015;10(Suppl). RO 6.1.
[91] Tierney JF, Fisher DJ, Burdett S, et al. Comparison of aggregate and individual participant data approaches to meta-analysis of randomised trials: an observational study. PLoS Med 2020;17(1): e1003019.
[92] Poggio F, Bruzzone M, Ceppi M, et al. Single-agent PARP in- hibitors for the treatment of patients with BRCA-mutated HER2- negative metastatic breast cancer: a systematic review and meta- analysis. ESMO Open 2018;3(4):e000361.AG-014699