REVIEWS Appraising iniparib, the PARP inhibitor that never was—what must we learn? Joaquin Mateo, Michael Ong, David S. P. Tan, Michael A. Gonzalez and Johann S. de Bono Abstract | Several drugs targeting poly(ADP-ribose) polymerase (PARP) enzymes are under development. Responses have been observed in patients with germline mutations in BRCA1 and BRCA2, with further data supporting antitumour activity of PARP inhibitors in sporadic ovarian cancer. Strategies to identify other predictive biomarkers remain under investigation. Iniparib was purported to be a PARP inhibitor that showed promising results in randomized phase II trials in patients with triple-negative breast cancer. Negative results from a phase III study in this disease setting, however, tempered enthusiasm for this agent. Recently, data from in vitro experiments suggest that iniparib is not only structurally distinct from other described PARP inhibitors, but is also a poor inhibitor of PARP activity. In this context, the negative iniparib phase III data might have erroneously promulgated the notion that PARP inhibition is not an effective therapeutic strategy. Here, we scrutinize the development of iniparib from preclinical studies to registration trials, and identify and discuss the pitfalls in the development of anticancer drugs to prevent future late-stage trial failures. Mateo, J. et al. Nat. Rev. Clin. Oncol. 10, 688–696 (2013); published online 15 October 2013; doi:10.1038/nrclinonc.2013.177
Introduction
Drug Development Unit, Division of Cancer Therapeutics and Division of Clinical Studies, The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, Downs Road, Sutton, Surrey SM2 5PT, UK (J. Mateo, M. Ong, D. S. P. Tan, M. A. Gonzalez, J. S. de Bono). Correspondence to: J. S. de Bono johann.de-bono@ icr.ac.uk
Over the past 5 years, around 4,800 early phase clini cal trials have been initiated in cancer medicine. During this period, more than 2,300 phase III studies were initi ated. Sadly, the rate of failure in oncology for novel com pounds undergoing clinical evaluation continues to be high compared with other medical specialties, such as cardiovascular diseases, and might even be rising with the advent of targeted therapies.1–3 Costs of drug develop ment also continue to increase, and may now be more than US$3 billion per drug approved; moreover, the market narrows the room for new approvals,4 impacting the price of compounds when approved and jeopard izing the sustainability of health services.5,6 Failures in late-phase registration clinical trials are especially disappointing, as they represent an enormous expendi ture of money and time from pharmaceutical companies and academic institutions, and signify potential suffering of huge numbers of patients with advanced-stage cancer and their families. Why are promising results observed in preclini cal research often not translated into clinical success? Several causes contribute to failures including: limited knowledge of cancer biology (despite ongoing and encouraging advances), selection of pharmacological compounds with suboptimal pharmacological proper ties, poorly predictive preclinical models, inappropriate trial designs, or decision-making based on nonrelevant end points.7–11 Iniparib (BiPar Sciences and Sanofi) is a compound that was initially developed as a poly(ADP–ribose) Competing interests The authors declare no competing interests.
688 | DECEMBER 2013 | VOLUME 10
polymerase (PARP) inhibitor, a class of drugs that impairs single-stranded DNA break repair. In general, PARP inhibitors have been useful in one of two strat egies: first, ‘synthetic lethality’, capitalizing on the sensi tivity of cells with defective homologous-recombination (HR)-mediated DNA repair, and second, sensitization of cells to DNA-damaging therapies. Indeed, several PARP inhibitors have demonstrated promising preclinical and clinical antitumour activity with wide therapeutic indices in the setting of tumours with BRCA1 or BRCA2 germ line inactivating mutations.12–15 Moreover, numerous preclinical models have also demonstrated that PARP inhibition can sensitize cells to the DNA-damaging effects of ionizing radiation, alkylating agents, and topoisomerase I targeting.16–19 Recent studies have raised concerns that iniparib is not a bona fide PARP inhibitor.20,21 The implications of these new data must be carefully considered, because over 2,500 patients have been treated in clinical trials of iniparib that were designed with PARP inhibition as a therapeutic goal (Table 1). These new data should caution against generalizing the negative results of iniparib trials to the ongoing development of other PARP inhibitors. In this Review, we retrace the development of iniparib from preclinical studies through to phase I–III studies and analyse lessons we must learn to optimize the development of other targeted drugs.
Preclinical studies Iniparib was developed as a prodrug of the more reac tive, but unstable, 4-iodo‑3-nitrosobenzamide (INOBA). INOBA inactivates PARP via two mechanisms: first, zinc-ejection following oxidation of the first zinc-finger
www.nature.com/nrclinonc © 2013 Macmillan Publishers Limited. All rights reserved
REVIEWS domain of the PARP protein resulting in loss of DNAstimulated PARP activity without loss of DNA binding capacity; second, induction of PARP-degrading amino peptidases that generate a characteristic polypeptide degradation product.22–25 Importantly, these purported mechanisms of action for iniparib differed signifi cantly from ‘classic’ PARP inhibitors, which specifi cally target the NAD+ binding site of PARP1 or PARP2, mimicking the NAD+ substrate by resembling the NAD+ moiety and competitively blocking PARP activity.26 Interestingly, the antitumour activity of INOBA also seemed to depend on the levels of glutathione and other reducing compounds. Depletion of glutathione in vitro enhanced antitumour activity and the lack of reducing flavoproteins in malignant cells, that convert INO2BA into nontoxic amines rather than to INOBA, was pro posed as a mechanism of selective tumoricidal action of INO2BA in cancer cell cultures, but not in fibroblast cultures. 27 Importantly, with or without glutathione depletion, in vitro experiments required substantial concent rations of INO 2BA for inhibition of PARP activity and antitumour activity. There are limited published preclinical studies of inipar ib either as a single agent or in combin ation with chemotherapy; antiproliferative effects have been reported in cancer cell lines including triple-negative breast cancer (TNBC; oestrogen [ER]-receptor-negative, progesterone [PR]-receptor-negative, and HER2negative) in which iniparib caused cell-cycle arrest in the G2/M phase.27,28 However, whereas synthetic lethal ity in BRCA-mutant cell lines was confirmed for other NAD+-competitive PARP inhibitors before clinical evalu ation,12,26 potent activity of iniparib related to selective inhibition of PARP in BRCA-deficient cultured cells has yet to be reported. Moreover, very few results were avail able within the public domain on the capacity of inipa rib to sensitize cells to DNA-damaging chemotherapy, including platinum agents,29,30 at the time the clinical trials started. In vitro studies, however, have suggested that iniparib does not exhibit the properties of a classic PARP inhibi tor. The effects of two different structural classes of NAD +-competitive PARP inhibitors (benzimidazole and pyridazinone derivatives) were compared with the effects of iniparib and its C‑nitroso metabolite (INOBA).26 The effects of these compounds were tested on BRCA1-deficient (MDA-MB‑436, exon 20 mutation), BRCA2-deficient (DLD1 –/–) and BRCA1/2-proficient (MDA-MB‑231 and DLD1+/+) breast cancer cell lines. All NAD+-competitive PARP inhibitors showed high selectivity for PARP, inhibited PARP enzymatic activ ity at nanomolar concentrations, inhibited autoribosy lation of PARP-1, potentiated alkylating chemotherapy (temozolomide), and showed selective activity in BRCAdeficient tumour cell lines and xenograft models. By con trast, iniparib and its metabolite were not able to inhibit PARP enzymatic activity, diminish poly(ADP–ribose) formation, potentiate temozolomide, or show activity in either BRCA-deficient or BRCA-proficient cell lines or xenograft models.26 The doses of iniparib required to
Key points ■■ Iniparib is not a bona fide inhibitor of poly(ADP-ribose) polymerase (PARP), so the clinical results in this context should not be extrapolated to other PARP inhibitors in development ■■ Preclinical data on iniparib did not sufficiently elucidate the mechanism of action of this agent before clinical trials were initiated ■■ Phase I trials should provide proof of mechanism and, ideally, proof of concept, in expansion cohorts to test biological hypotheses; early clinical trials of iniparib lacked proof of mechanism ■■ Selection of a patient population, and implementation and validation of predictive biomarkers, are critical to optimize drug development ■■ Randomized phase II trials have a significant rate of false positivity, so promising results should be interpreted prudently until other confirmatory studies are reported ■■ Preclinical and clinical studies with negative results and efforts evaluating reproducibility of previously published data should be publically available to minimize the risk of publication bias
achieve a cytotoxic effect were very high (>40 μmol/l). Furthermore, depletion of glutathione, a potential mediator of resistance, did not appreciably alter these observations. Instead of selective potent activity, inipa rib was found to nonspecifically react and form adducts with proteins containing cysteine residues, including the PARP‑1 zinc finger domain. In a separate study, three different BRCA-proficient TNBC cell lines were tested against iniparib and three different NAD+-competitive PARP inhibitors: AG‑014699 (rucaparib, Pfizer-Clovis), AZD‑2281 (olaparib, AstraZeneca) and ABT‑888 (veliparib, Abott Laboratories),21 showing lower potency of PARP inhibition and anti tumour activity for iniparib. BRCA1 knockdown sensi tized cells to iniparib, but it did not result in an increase of γ‑H2AX formation as occurred with rucaparib, olapa rib, and veliparib. An in vitro study of 12 breast cancer cell lines demonstrated higher IC50 concentrations for iniparib (13–70 μM) when compared with olaparib (IC50 range 3.7–31 μM).31 Another study exposed HR‑deficient cells (BRCA2deficient PEO1 human ovarian cancer cells and ATM‑deficient GM16666 fibroblasts) and HR‑proficient cells (BRCA2-revertant PEO4 ovarian cancer cells and ATM-restored GM16667 fibroblasts) to veliparib, olapa rib, and iniparib.32 The HR‑deficient cells were selectively sensitive to veliparib and olaparib; however, iniparib did not selectively target HR‑deficient cells. Furthermore, iniparib—unlike the other tested agents—did not synergize with either topoisomerase I poisons, cispla tin, gemcitabine, or paclitaxel in various cell lines, and failed to inhibit poly(ADP–ribose) formation even at concentrations of 100 μmol/l. Overall, these preclinical data suggest that the poten tial cytotoxic effects of iniparib are not mediated by PARP inhibition, but by mechanisms that are yet to be elucidated that might be related to stimulation of intra cellular production of reactive oxygen species.33 These findings became apparent only after the pursuit of a large and expensive drug development programme, and clearly question whether the preclinical evidence was strong enough to justify the initiation of clinical trials.
NATURE REVIEWS | CLINICAL ONCOLOGY
VOLUME 10 | DECEMBER 2013 | 689 © 2013 Macmillan Publishers Limited. All rights reserved
REVIEWS Table 1 | Clinical studies of iniparib as a single agent or combined with other drugs Trial initiation
Trial identifier
n
Phase
Tumour type
Treatment arms
March 200634
NCT00298675
58
I
ASC
Iniparib; iniparib + irinotecan
January 2007
NCT00422682
84
I
ASC
Iniparib + topotecan Iniparib + temozolamide Iniparib + gemcitabine Iniparib + carbo/paclitaxel
March 200848
NCT00687765
120*
I/II
Gliomas
Iniparib + temozolamide
July 2010
NCT01161836
7*
I
ASC
Iniparib (ECG-effect study)
September 2010
NCT01213381
18*
I
ASC
Iniparib + gemcitabine/carboplatin
November 2011
NCT01455532
160*
I/IB
ASC
Iniparib Iniparib + gemcitabine + carboplatin Iniparib + paclitaxel Iniparib + PEG-doxorubicin
September 2012
NCT01551680
30*
I
Brain M1
Iniparib + radiotherapy
October 200741
NCT00540358
123
II
TNBC
Iniparib + gemcitabine + carboplatin vs gemcitabine + carboplatin
December 200843
NCT00813956
80
II
TNBC (neoadjuvant)
Iniparib + gemcitabine + carboplatin
May 2008
NCT00687687
22
II
Uterine carcinosarcoma
Iniparib + carboplatin + paclitaxel
June 2008
NCT00677079
12*
II
BRCA1/2 ovarian
Iniparib
45
December 2009
NCT01033123
41
II
Ovarian (platinum-sensitive)
Iniparib + gemcitabine + carboplatin
December 200946
NCT01033292
43
II
Ovarian (platinum-resistant)
Iniparib + gemcitabine + carboplatin
February 2010
NCT01045304
163*
II
TNBC
Iniparib (weekly vs twice a week) + gemcitabine + carboplatin
May 201047
NCT01086254
116
II
NSCLC
Iniparib + gemcitabine + cisplatin vs gemcitabine + cisplatin
July 201083
NCT01173497
37
II
TNBC-brain M1
Iniparib + irinotecan
September 201042
NCT01204125
141
II
TNBC (neoadjuvant)
Iniparib + paclitaxel
July 200944
NCT00938652
519
III
TNBC
Iniparib + gemcitabine + carboplatin vs gemcitabine + carboplatin
March 201051
NCT01082549
780*
III
SCC lung
Iniparib + gemcitabine + carboplatin vs gemcitabine + carboplatin
NCT01593228
–
III
ASC
Iniparib extension programme
Phase I trials
36,37
Phase II trials
49
Phase III trials
Other May 2012
*Planned accrual as per ClinicalTrial.gov website; final results not reported. Abbreviations: ASC, advanced solid cancers; ECG, electrocardiogram; NSCLC, non‑small-cell lung cancer; SCC, squamous cell carcinoma; TNBC, triple‑negative breast cancer; vs, versus.
Clinical studies Phase I trials Single-agent iniparib was first evaluated in patients with advanced solid cancer in a phase I study with a ‘3 + 3’ dose-escalation design (NCT00298675) (Table 1).34 Results were reported for 24 patients treated at eight dose levels (0.5–8.0 mg/kg intravenous [IV] adminis tration per day, twice weekly). No dose-limiting toxici ties were observed and the most common adverse events were gastrointestinal disorders (39%). Best response of stable disease for ≥2 months was documented for six out of the 24 patients. Pharmacokinetic analysis showed rapid conversion of iniparib (half-life [T1/2] 4 min) into its active metabolites, but a later study reported that most of the drug is rapidly transformed into inactive com pounds.35 At a dose of 1.4 mg/kg, the maximum achieved concentration (Cmax) was above the threshold for efficacy 690 | DECEMBER 2013 | VOLUME 10
in preclinical models. Pharmacodynamic data showed inhibition of PARP in peripheral blood mononuclear cells (PBMCs) by >50% after a single dose of 2.8 mg/kg, and greater PARP inhibition after multiple doses.34 Studies of other biomarkers of PARP inhibition, such as induction of γ‑H2AX foci, were not reported and pharmacodynamic analysis in tumour tissue was not pursued. Unfortunately, most of the information about the assays used for the pharmacodynamic studies and their validity were not detailed at the time the results were communicated. What was defined as the ‘biologi cally relevant dose’ was estimated by pharmacokinetic and pharmacodynamic studies; moreover, a maximum tolerated dose (MTD) was never defined and specific expansion cohorts to prove target modulation in popu lations with known HR-mediated DNA repair system aberrations were not conducted.
www.nature.com/nrclinonc © 2013 Macmillan Publishers Limited. All rights reserved
REVIEWS A phase Ib study combined iniparib with four different chemotherapy regimens: topotecan (n = 14), gemcitabine (n = 22), temozolomide (n = 17), and carboplatin plus paclitaxel (n = 13) (NCT00422682).36 The doses of inipa rib explored ranged from 1.1–11.2 mg/kg (IV, days 1 and 4 of every week). Surprisingly, no dose-limiting toxicities were reported, and seven of 66 patients (10%) achieved radiological partial or complete responses as their best response to treatment. Further safety data from an expan sion cohort who received carboplatin (AUC6, day 1) plus paclitaxel (200 mg/m2, day 1) with iniparib (5.6 mg/kg, IV, days 1, 4, 8, and 11) again showed a low incidence of grade 3 or 4 adverse effects neutropenia (13.3%) and anaemia (6.7%).37 These data contrast with PARP inhibitor trials in which single-agent dose-escalation was limited by haematological toxicity, which was further potentiated by combination with chemotherapy.14,38,39
Phase II trials in TNBC An open-label, randomized phase II study in 123 patients with TNBC evaluated gemcitabine (1,000 mg/m2, days 1 and 8) plus carboplatin (AUC 2, days 1 and 8)40 with or without iniparib (initially 4.0 mg/kg, days 1, 4, 8, and 11; later amended to 5.6 mg/kg on the same schedule) once every 3 weeks (NCT00540358).41 The majority of patients (58.5%) had no prior systemic treatment for metastatic disease. The efficacy analyses evaluated the intentionto-treat population but, unlike the later phase III trial, radiological response assessments were performed by the investigators and were not subjected to an external blinded review. Patients with or without BRCA mutations were eligible, but the number of patients with BRCA mutations and their outcome have not been reported. The clinical benefit rate (CBR), which is the rate of com plete or partial radiological response together with stable disease >6 months, and objective response rate (ORR) were 34% and 32% in the chemotherapy alone arm, and 56% and 52% in the iniparib arm, respectively. The trial was designed to detect an improvement in CBR, assum ing a rate of 0.45 in the control group and expecting a CBR >0.60 in the experimental arm (with a power of 80% and accepting a two-sided α error of 0.05). Consequently, the study was considered positive (P = 0.001 for CBR; P = 0.002 for ORR). Secondary survival end points were also positive, but subject to very broad 95% confidence intervals; median progression-free survival (PFS) was 3.6 (2.6–5.2) months in the chemotherapy alone group versus 5.9 (4.5–7.2) months in the iniparib group (P = 0.01). Overall survival was 7.7 (6.5–13.3) months compared with 12.3 (9.8–21.5) months (P = 0.01) for the iniparib arm. Interestingly, no difference in the rate of adverse events, including haematological events, was observed between the two arms, again raising concerns regarding the lack of class-type toxicities associated with PARP inhibitors. A neoadjuvant trial of weekly paclitaxel (80 mg/m2, day 1) alone or with iniparib once weekly (11.2 mg/kg, IV, day 1) or iniparib twice weekly (5.6 mg/kg, IV) was also pursued in patients with TNBC (NCT01204125).42 Overall, 141 patients were recruited; the trial did not detect differences in the primary end point, rate of
pathological complete response (pCR) of the primary breast tumour. In another trial, 80 patients received 4–6 cycles of neoadjuvant gemcitabine (1,000 mg/m2, IV), carboplatin (AUC2, IV, days 1 and 8) and iniparib (5.6 mg/kg, IV, days 1, 4, 8, and 11) once every 3 weeks (NCT00813956). A pCR rate of 47% (90% CI 27–69%) was reported for the 19 carriers of BRCA mutation treated; for the BRCA wild-type participants, the pCR rate was 33% (90% CI 23–44%).43
Phase III trial in TNBC Iniparib quickly entered into phase III trials, labelled as a PARP inhibitor following the promising phase II data.41 A randomized open-label phase III study enrolled patients with TNBC who had undergone up to two prior lines of treatment.44 Patients were randomly assigned (1:1) to gemcitabine–carboplatin alone (gemcitabine 1,000 mg/m2, IV) and carboplatin (AUC 2; , IV, days 1 and 8) or the same regimen plus iniparib (5.6 mg/kg, IV, days 1, 4, 8, and 11) every 3 weeks. The trial was planned with two co-primary end points: overall survival and PFS. The sample size was calculated with the aim of detecting, with a power of 90%, a HR of 0.65 for PFS and 0.66 for overall survival between the two arms. The type 1 error accepted (0.05, two-sided) was divided for the two co-primary end points (0.04 for overall survival and 0.01 for PFS). The trial was to be considered posi tive if either one of the two primary end points was met. Between July 2009 and March 2010, 519 patients were enrolled and randomly assigned, which constituted overrecruitment of patients as the statistical design demanded 420 patients to test the hypothesis. Patient demographics suggested a similar population to the prior phase II trial. In total, 152 of 258 patients (59%) on the chemotherapy- only arm crossed over to receive chemotherapy and iniparib following disease progression. ORR was not significantly different (30% versus 34%), nor was the study positive for either of the co-primary end points.44 Clinical trials in other tumour types Two single-arm trials in both platinum-sensitive and platinum-resistant recurrent ovarian carcinoma have tested iniparib in combination with carboplatin and gem citabine (NCT01033123 and NCT01033292).45,46 A signifi cant number of BRCA-mutation carriers were enrolled, but no relationship between BRCA status and objective response was observed.14,38 Other trials have explored the antitumour activity of ini parib in combination with several chemotherapy regimens in lung cancer, gliomas, and uterine carcinosarcoma.47–49 Overall, none of these trials has provided either proof of concept of chemosensitization with iniparib or identi fied predictive biomarkers of response. A phase II trial in patients with lung cancers and all histologies but largely non-squamous did not report improvement of overall sur vival or PFS.47 In parallel, a large randomized phase III study in patients with newly diagnosed squamous-cell lung cancer was pursued (NCT01082549).50 Final results are yet to be presented, but a negative outcome has been announced in a press note (Table 1).51
NATURE REVIEWS | CLINICAL ONCOLOGY
VOLUME 10 | DECEMBER 2013 | 691 © 2013 Macmillan Publishers Limited. All rights reserved
REVIEWS Appraising iniparib: critical questions Was sufficient preclinical data acquired? When the early clinical trials began, limited published preclinical data supported the clinical evaluation of iniparib in patients with cancer. Iniparib was ‘labelled’ as a potent and selective PARP inhibitor, and clinical trials were designed to recruit patients with presumed DNA repair deficiency. Later attempts to evaluate inipa rib by independent investigators failed to find supporting evidence of potent and selective PARP inhibition. This inability to validate the mechanism of action, the lack of in vitro data supporting a selective effect on HR‑deficient cell lines, and the lack of robust evidence of synergy in combination with chemotherapy raises significant con cerns that the preclinical evidence was insufficient to initiate clinical studies. Reproducibility of research results is an often under estimated concern, at least in part because of academic and/or industry pressures to accelerate drug develop ment programmes, maximize fiscal return, and achieve academic recognition. When selecting a compound to enter clinical evaluation, careful scrutiny is needed of the available preclinical data, the validity of the assays used, and the robustness of the results to justify the high amount of personal and monetary resources to be employed. Two independent studies evaluating the reproducibility of results published by either academic researchers or pharmaceutical industry investigators in high-impact publications have recently exposed the problem: only 11–25% of published results could be replicated by two groups of industry investigators, even when using identical reagents (usually provided by the original laboratories).52,53 The issue is also relevant when developing drug combinations, as the concept of ‘synergy’ is often overestimated and clinical trials too often fail to confirm the expectations derived from preclinical results.54 Investigators, industry, and editors should also be encouraged to report and accept for publication ‘nega tive’ preclinical studies to limit such publication bias, and to await a robust confirmation and appraisal of preclinical data before clinical trials are commenced. Were proof of mechanism and concept pursued? Given the extremely short reported T1/2 of iniparib (4–11 min),34 a major concern is whether systemic expo sure to iniparib or its active metabolites are significant enough for clinical activity. Although Cmax plasma levels of iniparib exceeded concentrations necessary for in vitro activity, few data are available regarding plasma expo sure (such as AUC) necessary for antitumour activity. Although early phase trials reported a T1/2 of 2–4 h for the active metabolites, Verweij et al.35 presented contradicting data that T1/2 of metabolites was