Table 1 lists the characteristics of study patients. In total, 46 patients were consented at 12 institutions between 7/2010 and 8/2012. Nine patients were consent-not-treat (n = 7 with PD during screening, n = 1 with Gilbert’s Syndrome, n = 1 with hemorrhage on baseline brain MRI). Thirty-seven patients were evaluable for toxicity and 34 for efficacy. Of the 24 patients for whom BRCA status was known, 21 % harbored a germline mutation (4 BRCA1, 1 BRCA2). Recursive partitioning analysis (RPA) score was calculated and was 1, 2 and 3 in 19, 59 and 22 % of patients, respectively .
Toxicity and dose Intensity
Irinotecan plus iniparib was generally well-tolerated (n = 37, Fig. 1). The most common grade 1/2 adverse events were diarrhea (51 %), nausea (46 %), fatigue (32 %), neutropenia (22 %), and constipation (22 %). The most common grade 3 or greater adverse events were neutropenia (14 %), leukopenia (5 %) hypokalemia (5 %), and decreased white blood cells (5 %).
Disease progression as defined by radiographic imaging was the most common reason for treatment discontinuation (27/34, 79 %). Four patients (12 %) experienced functional decline in the absence of radiographic progression. No patient had a serious adverse event that led to study withdrawal. One patient died while on protocol due to overwhelming respiratory infection unrelated to treatment, one withdrew for personal reasons, and one remained on protocol therapy at the time of analysis.
Eight of 34 patients evaluable for efficacy (24 %) required an initial dose modification of irinotecan to 100 mg/m2, and 3 of 8 required a second dose modification to 75 mg/m2. Seventeen (50 %) required a dose delay due to hematologic toxicity. Of the five patients who escalated the dose of iniparib to 8 mg/kg mid-study, one patient de-escalated back down to 5.6 mg/kg due to grade 2 nausea. There was no difference between grade 3/4 adverse events by dose of iniparib (5.6 mg/kg vs. ever receiving 8 mg/kg, p = 0.27) or by BRCA mutation status (p = 0.99).
Time to progression
Among 34 evaluable patients, 32 (94 %) experienced either intracranial or extracranial progression (or both), one died on study, and the other remained on treatment as part of an extension study. The median TTP was 2.14 months (95 % CI 1.74–4.34, Fig. 2a). In exploratory analysis, there was a significant improvement in TTP by iniparib dose (p = 0.009). Median TTP was 7.8 months for 9 patients who received 8 mg/kg, 5.7 months for 5 who received both doses, and 1.9 months for 23 who received 5.6 mg/kg. There was no difference in TTP by BRCA mutation status (2.3 months for BRCA wild-type vs. 1.6 months for BRCA mutation carriers, p = 0.91).
At the time of the analysis, 27 (79 %) of patients had died. Median OS was 7.83 months (95 % CI 5.10–10.2, Fig. 2b). There was no difference in OS by either iniparib dosing (p = 0.24) or BRCA mutation status (p = 0.49).
Objective response rate
Of 34 evaluable patients, the intracranial response rate was 12 % (0 CR, 4 PR’s, Table 2). Thirteen additional patients (41 %) had SD as best response intracranially, while for 15 (47 %), it was PD. The intracranial CBR, including those with PR and SD for ≥6 months, was 27 % (9/34). Of the n = 4 patients who achieved an intracranial PR, 2 harbored a BRCA1 mutation. Extracranial disease was evaluable in 19 patients, of whom 1 had a PR (5 %). Six patients (32 %) experienced extracranial SD, while 12 (63 %) experienced PD as best response. The extracranial CBR was 11 % (2/19). Site of first progression was intracranial in 38 %, extracranial in 31 % and both intra- and extracranial in 31 %.
Volumetric analysis of intracranial response
Intracranial response was evaluated using uni-dimensional (modified RECIST)  and three-dimensional measurements (volumetric criteria) . Intracranial PR by volumetric analysis (≥50 % reduction) was observed in 6 patients (22 %), SD in 11 (41 %) and PD (≥40 % increase) in 10 (37 %) (Fig. 3).
Comparing intracranial response by modified RECIST and volumetric criteria, discordance in response classification was observed (10/27, 37 %). It was more common for volumetric response to overestimate response when compared to uni-dimensinoal response classification (7/10, 70 %), see Supplemental Table 1 and Supplemental Results. Achieving a PR by RECIST criteria was significantly associated with OS (HR = 0.12, 95 % CI 0.015–0.877, p = 0.037), while a PR as determined by volumetric criteria did not (HR = 0.41, 95 % CI 0.120–1.39, p = 0.15).
Health-related quality of Life (HRQL)
Of 37 patients evaluable for toxicity, 11 completed the baseline HRQL only, and 25 completed at least a baseline and one follow-up HRQL. For details on participants, see Supplemental Results. To provide a detailed picture of subjects’ HRQL, FACT subscales were all analyzed separately. For the entire population, the only significant change was in physical well-being, which declined from pre-study to first follow-up HRQL assessment (p < 0.01). We compared the change in HRQL from pre-study to first follow-up between subjects who received a clinical benefit from the treatment versus those who did not and found no significant difference, Supplemental Table 2.
Intrinsic subtyping and correlative endpoints
FFPE tumor was available from 21 patients, of which 24 total samples were analyzed (17 primary breast tumors, 2 lymph nodes, 2 brain metastases and 3 lung metastases). Three were matched pairs; 2 breast primary/lymph node pairs, 1 breast primary/lung metastases from the same patient.
Intrinsic molecular subtyping identified 19 (79 %) basal-like, 2 (8 %) HER2-enriched, and 3 (13 %) normal-like tumors. Concordant with a previous molecular characterization of TNBC , the two HER2-enriched tumor samples did not show high ERBB2 gene expression, but showed high EGFR gene expression. Notably, no luminal A, B or claudin-low tumors were identified. Subtype concordance between the 3 matched pairs was 30 % with 1 breast/lymph node pair both basal-like, 1 breast/lung metastases pair basal-like/normal-like and 1 breast/lymph node pair basal-like/HER2-enriched.
We evaluated the association between expression of 7 gene signatures (as a continuous variable) and intracranial response (PR vs. SD/PD). High expression of proliferation and ROR-P scores were significantly associated (p = 0.046, Mann–Whitney test) with a PR (Fig. 4). When the previously defined cutoffs to define low/medium/high ROR-P expression were evaluated , 4/10 patients with tumors classified as ROR-P high had an intracranial PR versus 0/9 patients with tumors classified as ROR-P medium (unilateral p = 0.0576, Fisher’s exact test). No tumor was classified as RORP-low. There was no association between response and the VEGF-13 gene signature.
Secondly, we evaluated the association between expression of 123 breast cancer-associated candidate genes, and 5 house-keeping genes (as a continuous variable) and intracranial response (PR vs. SD/PD; PR vs. PD and PR vs. SD) (Supplemental Data 2). The comparison that yielded the largest number of significant genes (n = 38, 30.9 %) was PR vs. PD [false discovery rate (FDR) <30 %]. Concordant with the previous findings using gene signatures, the 13 genes (e.g. NDC80 and EXO1) whose high expression was associated with an intracranial PR were found to be enriched for cell cycle-related biological processes (p < 0.0001) (Supplemental Fig. 1). Conversely, the list of 25 genes whose high expression was associated with intracranial PD was found to be enriched for regulation of cytoskeleton-related (e.g. CAV1 and MAPT) and transcription-related (e.g. FOXA1 and RB) biological processes. Finally, the other two possible comparisons (i.e. PR vs. SD/PD and PR vs. SD) only yielded genes whose high expression was associated with lack of PR (FDR < 30 %). Interestingly, both of these gene lists were enriched for genes typically associated with HER2-enriched biology (e.g. FGFR4 and TMEM45B) and luminal biology (e.g. INPP4B, FOXA1, NAT1 and PGR).