Globo H allyl glycoside is a hexasaccharide prepared according to the total synthesis route, previously developed by Professor SJ Danishefsky.19 The purity of Globo H allyl glycoside is ≥99%. OBI-821 is directly purified from the Quil-A, a saponin extract obtained from the tree bark of Quillaja saponaria, Molina. The purification steps involved three stages of chromatographic purification, developed by Optimer Pharmaceuticals (Jersey City, New Jersey, USA), from Quil-A to obtain the OBI-821 containing two major isomers and four other potential isomers as a mixture. Of these six isomers, four have been identified to be structurally identical to the currently marketed adjuvant, QS-21. The total purity of these six isomer mixtures is controlled at no less than 99%.

Eligible patients were women with MBC achieving SD, partial response (PR), or complete response (CR) after at least one anticancer therapy and with no more than two events of progressive disease after MBC diagnosis. Patients with estrogen receptor-positive (ER+) or progesterone receptor-positive (PR+) tumors were allowed to continue antihormonal therapy with study treatment; those completing chemotherapy and starting hormone therapy during maintenance therapy must have had SD for at least 4 weeks before study entry. Concurrent treatment with anti-human epidermal growth factor receptor 2 (HER2) therapies was prohibited. Patients had adequate organ function and an Eastern Cooperative Oncology Group (ECOG) performance status of less than or equal to 1. Exclusion criteria included more than two lines of prior anticancer therapy; chemotherapy within 4 weeks of randomization; autoimmune disease or disorder requiring treatment with systemic corticosteroids or immunosuppressive therapies; any other investigational drug; any evidence or history of central nervous system metastases; and bone-only metastases.

Eligible patients were randomly assigned to receive AS/OBI-821 or placebo (phosphate-buffered saline) in a 2:1 ratio via a centralized interactive web-based randomization system. Adagloxad simolenin (OBI-822) was used with OBI-821, a saponin-based adjuvant that contains the same major components as QS-21 used in the phase I trials.20

Patients were stratified according to disease status at randomization (CR or PR/SD) and hormone therapy use (yes or no), with a block size of three for each of the four combined stratification levels. Study subjects and investigators were blinded to treatment group assignment.

AS/OBI-821 (30 µg/100 µg) or placebo was administered by subcutaneous injection at weeks 1, 2, 3, 5, 9, 13, 17, 25, and 37 for a total of nine doses or until disease progression. All patients received cyclophosphamide 300 mg/m² administered intravenously at weeks 1, 5, 9, 13, 17, 25, and 37, 3 days prior to each dose of study drug.

CT scan (or MRI) was performed within 3 weeks before randomization, then repeated every 8 weeks for up to 2 years or until disease progression (whichever came first), reviewed both locally and by an independent central radiology facility. Biochemical and hematologic laboratory tests were performed within 3 weeks before randomization, then repeated at weeks 3, 5, 9, 13, 17, and every 8 weeks thereafter for up to 2 years or until disease progression or early termination. Blood samples were collected for measurement of anti-Globo H IgG and IgM titers by ELISA at weeks 1, 3, 5, 9, 13, 17, 25, 33, 37, 41, and every 8 weeks thereafter for up to 2 years or until disease progression or early termination. Anti-KLH IgG titers were assessed by ELISA in blood samples obtained at week 40. Blood samples for cellular immune response were collected on the day of but prior to dosing, then at day 4, weeks 5, 13, and 41. Additional immune studies were performed to assess antisera binding ability with Globo H+ tumor cells and CDC/ADCC, and results will be published elsewhere.

Archival samples (n=244) for Globo H analysis were from the primary tumor (184/244, 75%) or from metastatic sites (60/244, 25%). Metastatic sites included lymph nodes, chest walls, lung, liver, ovary, bone, and others. Archival tumor tissue was tested for Globo H antigen by immunohistochemical (IHC) staining with anti-Globo H IgG, VK9 monoclonal antibody (provided by Memorial Sloan Kettering Cancer Center, New York, USA) as previously described.21 Globo H is a glycolipid molecule. In this study, the interval between sample collection and IHC assay ranged from 0 to 17.3 years, with median of 1.5 years. The Globo H scores of the ≤10-year-old samples were comparable with the scores of the >10-year-old samples (p=0.23). We compared Globo H expression in 18 available pairs of primary tumors versus metastatic lesions from the same patients collected more than 1 year apart. The median duration of sample collection between primary and metastatic tumors was 2.7 years (range, 1.3–11.7 years). There was no obvious correlation of Globo H expression between primary and metastatic tumors obtained at an interval of more than 1 year. Among 14 patients whose paired tumors were collected at the same time (n=10) or at 1-month to 6-month intervals (n=4), there also was no correlation of Globo H expression between primary and metastatic tumors. In all paired samples obtained from 32 patients, there was no significant correlation of Globo H expression between primary and metastatic tumors. Thus, long-term storage of paraffin-embedded tissues of up to approximately 17 years did not appear to affect the detection of Globo H expression by IHC. Based on the percentage of tumor cells with positive staining, Globo H expression was defined as negative (<1%), 1+ (1%–30%), 2+ (31%–60%), or 3+ (>60%). Adverse events (AEs) were recorded and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events V.4.03, with relationship to study medications recorded.

The primary endpoint was investigator-assessed PFS, which was defined as the time from randomization to investigator-assessed progression or death, whichever came first. Secondary endpoints included OS, defined as the interval from randomization until death from any cause; magnitude and correlation of the humoral immune response (IgM and IgG against Globo H) with PFS and OS; correlation of tumor Globo H expression with study outcome; and safety. Post hoc analyses included an analysis of PFS in patients who completed all nine injections of study drug; an analysis of anti-KLH IgG levels at week 40 (when all nine injections of study drug were scheduled to have been administered) in patients who did and did not have a humoral immune response to the investigational product (retrospectively defined as an anti-Globo H IgG titer ≥1:160 and <1:160, based on the correlation of anti-Globo H IgG response with PFS); and an analysis of potential clinical factors associated with anti-Globo H IgG response (retrospectively defined as an anti-Globo H IgG titer ≥1:160).

Efficacy and baseline variables were evaluated in the modified intent-to-treat (mITT) population, which included all patients who received at least one injection of study drug and were analyzed according to randomized treatment. The safety population included all patients who received at least one injection of study drug and who had at least one post-dose safety assessment and were analyzed according to treatment received.

PFS and OS for each treatment group were evaluated using the Kaplan–Meier method with treatment effect and treatment comparison (HR, associated CI, and p value) estimated from the Cox proportional hazards model stratified by hormone therapy use during study treatment (yes or no) and investigator-reported baseline disease status (CR or PR/SD). The proportional hazards model was also used to assess the potential factors associated with the immune response and the post hoc analyses on PFS. Safety variables were analyzed with treatment comparison using Fisher’s exact test. All p values were based on two-sided tests with no multiple comparison adjustments. A total of 289 disease progression events or deaths were estimated to detect an HR of 0.67 with 90% power using a two-sided log-rank test at 5% level of significance. Assuming an extension of median PFS from 6 to 9 months (ie, HR 0.67), a 2-year recruitment and a 2-year follow-up period, and accounting for an 11% dropout rate, a sample size of 342 patients was required.

Between January 13, 2011, and August 25, 2014, a total of 349 patients were enrolled and randomly assigned to either AS/OBI-821 (n=225) or placebo (n=124), both in combination with cyclophosphamide. One patient randomized to AS/OBI-821 did not receive study treatment and was excluded from the mITT and safety populations (figure 1). Baseline characteristics were well balanced between the treatment groups (table 1). A total of 121 (75%) of the 161 patients in the AS/OBI-821 group and 63 (75%) of the 84 patients in the placebo group with ER+ or PR+ tumors received concurrent antihormonal therapy throughout the study.

The data cutoff date was November 1, 2015. Of the patients who received study treatment, 104 (46%) of 224 assigned to AS/OBI-821 and 64 (52%) of the 124 assigned to placebo received all nine scheduled study drug injections. The main reason for treatment discontinuation was disease progression (109 patients (49%) in the AS/OBI-821 group and 50 patients (40%) in the placebo group). Other causes included consent withdrawal (7 (3%) and 8 (6%), respectively), AEs (2 (<1%) and 2 (2%), respectively), non-compliance (one patient (<1%)) in the AS/OBI-821, and loss to follow-up (one patient (<1%)) in the AS/OBI-821 group).

At the time of this analysis, the median follow-up time for patients was 22.3 months in the AS/OBI-821 group and 21.2 months in the placebo group. A total of 159 patients (71%) in the AS/OBI-821 group experienced an investigator-assessed PFS event (all disease progression events) compared with 90 patients (73%) in the placebo group (89 disease progression events and one death). Median PFS by investigator assessment was 7.6 months (95% CI: 6.5–10.9) in the AS/OBI-821 arm and 9.2 months (95% CI: 7.3–11.3) in the placebo arm (HR=0.96; 95% CI: 0.74–1.25; p=0.77; figure 2). Exploratory analyses evaluating subgroups of patients based on stratification factors (hormone therapy use and disease status at baseline) and breast tumor subtypes (hormone receptor-positive (HR+)/HER2-negative (HER2–), triple-negative breast cancer (TNBC), HER2-positive (HER2+)) revealed no difference in PFS by treatment group within each subgroup (online supplementary figure 1).

At the time of the analysis for the PFS endpoint, 85 deaths were reported. These occurred in 50 patients (22.3%) in the AS/OBI-821 arm and 35 patients (28.2%) in the placebo arm. The majority of deaths were due to disease progression. The median OS had not been reached for either treatment group. Preliminary OS analysis of patients treated with AS/OBI-821 versus placebo showed an HR of 0.79 (95% CI: 0.51–1.22; p=0.29).

Of the 224 patients treated with AS/OBI-821, 199 (89%) generated a detectable anti-Globo H IgM (titer ≥1:20) and 182 (81%) generated a detectable anti-Globo H IgG (titer ≥1:20) at least once during the study. An evaluation of the correlation between anti-Globo H IgG antibody titer levels and PFS outcome revealed that the higher the anti-Globo H IgG antibody levels, the better the PFS outcome, and the curve with a titer level of 1:160 started showing a trend of better PFS than the placebo curve (figure 3A). Using an anti-Globo H IgG titer of 1:160 as a cutoff, patients treated with AS/OBI-821 with an IgG titer ≥1:160 at any time during study treatment (n=112) had improved median PFS (11.1 months (95% CI: 9.3–17.6)) compared with those who never achieved an IgG titer ≥1:160 (n=112) (5.5 months (95% CI: 3.7–5.6)); HR=0.52; 95% CI: 0.37–0.71; p<0.0001, and placebo-treated patients (n=124) (9.2 months (95% CI: 7.3–11.3)); HR=0.69; 95% CI: 0.50–0.94; p=0.02 (figure 3B). In general, AS/OBI-821 was immunogenic, and IgM response occurred earlier than IgG response with median time to anti-Globo H titer 1:160 of 2.9 months (95% CI: 2.0–7.5) and 5.7 months (95% CI: 5.6–9.3) for IgM and IgG, respectively (figure 4).

In a post hoc analysis, KLH IgG levels at week 40 were available for 95 patients in the AS/OBI-821 arm and 59 patients in the placebo arm. As expected, patients who received AS/OBI-821 had a higher KLH IgG titer than those who received placebo; however, no significant differences in KLH IgG levels were observed between AS/OBI-821 recipients who had an anti-Globo H IgG titer ≥1:160 and those who had an anti-Globo H IgG titer <1:160 (online supplementary figure 2).

Of the 348 patients who received at least one dose of study treatment, tumor samples for Globo H antigen testing were available from 243 patients (70%); 159 in the AS/OBI-821 arm and 84 in the placebo arm (table 1). Globo H expression was evaluated and scored as 0 (AS/OBI-821, 44 vs placebo, 25), 1+ (57 vs 30), and 2+ and 3+ combined (58 vs 29) by IHC and was detected (Globo H expression 1+, 2+, or 3+) in 72% (115/159) of the AS/OBI-821-treated patients and in 70% (59/84) of the placebo-treated patients. For patients with no Globo H expression detected, the comparison for PFS between the AS/OBI-821 and placebo groups showed an HR of 0.75 (95% CI: 0.41–1.36; p=0.34). Similar analyses in the subgroups of patients with Globo H expression 1+ and with Globo H expression 2+ and 3+ combined showed HRs of 1.22 (95% CI: 0.72–2.07; p=0.46) and 0.77 (95% CI: 0.46–1.30; p=0.33), respectively. This analysis evaluating the potential correlation between percentage of Globo H-expressing cells and PFS was inconclusive.

Post hoc analyses suggested that a greater percentage of patients without progressive disease events after diagnosis of metastatic disease had an anti-Globo H IgG ≥1:160 (57% (78/137)) compared with those with one or two events of progressive disease (39% (34/87)). In addition, a higher percentage of patients with CR status at the time of randomization had anti-Globo H IgG ≥1:160 (62% (18/29)) compared with those with non-CR status (PR or SD) at baseline (48% (94/195)). A higher percentage of patients with ER+ disease achieved anti-Globo H IgG ≥1:160 (53% (95/180)) compared with those with ER− disease (39% (17/44)). However, given the relatively small sample size, the significance of these associations awaits validation in future studies.

Globo H expression was analyzed by breast cancer tumor subtypes (TNBC, HER2+, HR+). The percentages of patients with 0, 1+, 2+, and 3+ were not different by cancer type (online supplementary table 1).

The baseline characteristics between patients receiving all nine injections and those who received less than the nine injections were similar (online supplementary table 2). In the patients who had completed all nine injections of study drug, median PFS was 20.7 months (95% CI: 18.5−not estimable) in the AS/OBI-821 arm and 16.7 months (95% CI: 13.3−22.3) in the placebo arm (HR=0.66; 95% CI: 0.42−1.01; p=0.06) (figure 5).

A similar incidence of all treatment-emergent adverse events (TEAEs) (98.2% vs 96.0%) and non-injection site TEAEs (95.1% vs 94.4%) occurred in the AS/OBI-821 and placebo arms, respectively (table 2). TEAEs at the injection site were more common in the AS/OBI-821 arm than in the placebo arm (77.2% vs 15.3%; p<0.0001). All injection site reactions were grade 1 or 2 in severity and transient, and the incidence decreased with successive injections. A statistically significant higher incidence of fever was reported with AS/OBI-821 (20.1%) compared with placebo (6.5%; p<0.0005). Most non-injection site TEAEs were grade 1 or 2 in severity.

Serious adverse events (SAEs) were reported in 29 patients (12.9%) in the AS/OBI-821 arm and in 15 patients (12.1%) in the placebo arm. There were three serious adverse drug reactions with AS/OBI-821 (hypersensitivity in two patients and fever in one patient; all recovered without sequelae).

Deaths related to SAEs occurred in four patients in the AS/OBI-821 arm (three due to disease progression and one due to central nervous system metastases) and in no patients in the placebo arm.

The first 40 patients who completed all nine planned injections and developed median or greater titers of IgM or IgG anti-Globo H antibodies (≥1:160) were selected and exploratory FACS, CDC, and ADCC assays were performed on immune sera to explore tumor cell binding and potential functional impact of these vaccine-induced polyclonal antibodies. Anti-Globo H IgM antibodies bound to Globo H-expressing MCF-7 cells had a peak 3.26-fold increase at 4 weeks, which was associated with a 1.66-fold peak increase in CDC also at 4 weeks (online supplementary figure 3A). Anti-Globo H IgG antibody binding to MCF-7 cells was not observed as a mean increase over baseline (online supplementary figure 3B), although sera from 17 of 40 patients tested demonstrated increases in ADCC over baseline of between 1.1-fold and 2.6-fold (online supplementary figure 4 and online supplementary table 3).