Background: Nadunolimab is a first-in-class, antibody-dependent cell-mediated cytotoxicity enhanced, monoclonal antibody targeting IL1 Receptor Accessory Protein (IL1RAP) on cancer cells, cancer stromal cells, and tumor infiltrating immune cells. Nadunolimab blocks both IL1α and IL1β signaling which is linked to tumor progression, therapy resistance, and immune suppression signals. The TRIFOUR trial (NCT05181462) is a Phase (Ph) Ib/II study evaluating nadunolimab in combination with chemotherapy as a 1L or 2L therapy in advanced triple negative breast cancer (aTNBC) patients. In the TRIFOUR PhIb part, nadunolimab was administered at either 1 mg/kg (n=3) or 2.5 mg/kg (n=12) in combination with gemcitabine (1000 mg/m2) and carboplatin (AUC 2 mg/mL/min) on days 1 and 8 of each 3-week cycle. Here we present updated safety and efficacy data along with emerging translational research results from the 15 PhIb patients and a characterization of aTNBC. Methods: PhIb, serum and blood samples were collected pre-treatment and during the study and analyzed for soluble biomarkers and blood immune cell populations by ELISA, the Olink Immuno-oncology 92plex protein panel and hematology. In a different set of aTNBC patients with ≤ 1 previous line of therapy for locally advanced or metastatic BC, biopsies (n=22) were characterized for expression of IL1RAP and IL1α by immunohistochemistry and blood cells were characterized for IL1RAP expression by flow cytometry (n=31). Results: The PhIb safety profile was comparable to historic control data for gemcitabine plus carboplatin alone. Grade ≥3 treatment-emergent adverse events were reported in 12 (80%) patients, leading to treatment discontinuation in 1 (7%) patient. Five (33%) patients experienced serious AEs, febrile neutropenia being the most frequent (2 [13%] patients). Updated overall response rate was 60% (95% CI: 32-84), median progression-free survival was 6.2 months (95% CI: 3.7-8.3) and median overall survival was 12.8 months (95% CI: 8.5-NE). No PhIb patients remain on treatment. Emerging translational analyses comparing C1D1 and C2D1 time points identified a decreased absolute neutrophil count (ANC) (Hodges-Lehmann estimate and confidence interval for difference (HLE), -1.33, 95% CI: -2.2 to -0.73, P=.003) and neutrophil to lymphocyte ratio (HLE, -1.0, 95% CI: -2 to -0.4, P=.005) as well as decreased C reactive protein (HLE, -6.0, 95% CI: -19.2 to -0.4, P=.001). IL8 was also decreased on treatment, with a trend of correlation to a better outcome. Characterization in the separate set of patients showed widespread IL1RAP expression on tumor and stromal cells in all biopsies while IL1RAP-positive immune cells were found in 20 out of the 22 (91%) biopsies. IL1α expressing tumor cells were present in all biopsies, as were IL1α positive immune cells. All blood samples characterized by flow cytometry contained IL1RAP-positive immune cells with IL1RAP expression being notably high on monocytic myeloid-derived suppressor cells. Conclusions: Data from the PhIb TRIFOUR study indicate that nadunolimab at 2.5mg/kg, combined with gemcitabine plus carboplatin, has acceptable safety and tolerability and shows promising antitumor activity. Analyses of immune cell subsets and biomarkers showed potentially beneficial effects on cells and markers related to inflammation and immune response, such as decreased ANC and IL8. Characterization of aTNBC affirmed IL1RAP expression on tumor cells, cancer fibroblasts as well as tumor-associated and blood immune cells. In the biopsies, IL1α was expressed both in cancer and immune cells. The randomized PhII part of the trial is currently enrolling patients at 2.5mg/kg and more translational analyses are underway. Citation Format: Marta Santisteban Eslava, Agostina Stradella, Silvia Antolín Novoa, Pablo Tolosa, Javier García Corbacho, Angel Guerrero-Zotano, Manuel Ruíz Borrego, Irati Garmendia, Paloma Petit de Prado, Juan José Soto-Castillo, Cristina Reboredo, Manuel Alva, Elin Jaensson Gyllenbäck, Petter Skoog, Nedjad Losic, Ignacio Garcia-Ribas, Maribel Casas, Isabel Romero-Camarero, Rosalía Caballero, Sara López-Tarruella Cobo, Susana Bezares, María Muñoz Caffarel. Safety, efficacy and emerging biomarker data from the Phase Ib part of a Phase Ib/II clinical study of nadunolimab in combination with gemcitabine and carboplatin in patients with advanced triple negative breast cancer (TRIFOUR study) [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P1-09-29.

Supplementary Figure 1 from Connexin-26 Is a Key Factor Mediating Gemcitabine Bystander Effect

Supplementary Figure Legend from Connexin-26 Is a Key Factor Mediating Gemcitabine Bystander Effect

(A) Best percent change from baseline in tumor size for all patients in schedule A. (B) and (C) computed tomography scans from patients in schedule A at screening (left panels) and 9 weeks post-dose (right panels); (B) patient with duodenal cancer treated with TAK-931 30 mg; (C) patient with esophageal cancer treated with TAK-931 50 mg. Yellow arrows indicate target lesions.

Abstract Gemcitabine is a nucleoside analogue with anticancer activity. Inside the cell, it is sequentially phosphorylated to generate the active drug. Phosphorylated nucleoside analogues have been shown to traffic through gap junctions. We investigated the participation of gap junctional intercellular communication (GJIC) as a possible mechanism spreading gemcitabine cytotoxicity in pancreatic tumors. Immunohistochemical analysis of pancreatic cancer biopsies revealed increased connexin 26 (Cx26) content but loss of connexins 32 (Cx32) and 43 (Cx43) expression. Cx26 abundance in neoplastic areas was confirmed by Cx26 mRNA in situ hybridization. Heterogeneity on the expression levels and the localization of Cx26, Cx32, and Cx43 were identified in pancreatic cancer cells and found to be associated with the extent of GJIC, and correlated with gemcitabine bystander cytotoxic effect. The abundance of Cx26 at the contact points in tumoral regions prompted us to study the involvement of Cx26 in the GJIC of gemcitabine toxic metabolites and their influence on the antitumoral effects of gemcitabine. Knockdown of Cx26 led to decreased GJIC and reduced gemcitabine bystander killing whereas overexpression of Cx26 triggered increased GJIC and enhanced the gemcitabine cytotoxic bystander effect. Gemcitabine treatment of mice bearing tumors, with a high GJIC capacity, resulted in a significant delay in tumor progression. Interestingly, gemcitabine administration in mice bearing tumors that overexpress Cx26 triggered a dramatic tumor regression of 50% from the initial volume. This study shows that Cx26 participates in the gap junction–mediated bystander cytoxic effect of gemcitabine and provides evidence that upregulation of Cx26 improves gemcitabine anticancer efficacy. Mol Cancer Ther; 10(3); 505–17. ©2011 AACR.

Purpose: LY2334737 is an orally available prodrug of gemcitabine. The objective of this study was to determine the maximum tolerated dose (MTD) and dose limiting toxicities (DLT) of daily administration of LY2334737 with or without erlotinib. Experimental Design: Patients with advanced or metastatic cancer were treated with escalating doses of LY2334737 monotherapy or in combination with continuous daily administration of 100 mg erlotinib. LY2334737 was given once daily for 14 days of a 21-day cycle. The study was extended with a bioequivalence trial to investigate a novel LY2334737 drug formulation. Results: A total of 65 patients were treated in this study. The MTD was 40 mg LY2334737. Fatigue was the most frequent DLT for LY2334737 monotherapy (4 patients) followed by elevated transaminase levels (2 patients), both observed at the 40- to 50-mg dose levels. Among the 10 patients in the combination arm, 2 had DLTs at the 40-mg dose level. These were fatigue and elevated liver enzyme levels. The most common adverse events were fatigue (n = 38), nausea (n = 27), vomiting (n = 24), diarrhea (n = 23), anorexia (n = 20), pyrexia (n = 18), and elevated transaminase levels (n = 14). The pharmacokinetics showed dose proportional increase in LY2334737 and gemcitabine exposure. The metabolite 2′,2′-difluorodeoxyuridine accumulated with an accumulation index of 4.3 (coefficient of variation: 20%). In one patient, complete response in prostate-specific antigen was observed for 4 cycles, and stable disease was achieved in 22 patients overall. Pharmacokinetic analysis showed that the 2 investigated LY2334737 drug formulations were bioequivalent. Conclusions: LY2334737 displays linear pharmacokinetics and the MTD is 40 mg with or without daily administration of 100 mg erlotinib. Signs of antitumor activity warrant further development. Clin Cancer Res; 17(18); 6071–82. ©2011 AACR.

Abstract Background: LY2603618 is a selective inhibitor of Chk1, a protein kinase that plays a key role in the DNA damage checkpoint. Inhibition of Chk1 is predicted to enhance the effects of antimetabolites, such as gemcitabine. Methods: This study is a Phase 1–2 study in patients with solid tumors (Phase 1) and advanced pancreatic adenocarcinoma (Phase 2). In Phase 1, gemcitabine (1000 mg/m2) was administered on Days 1, 8, and 15 of a 28-day cycle. LY2603618 was administered on Days 2, 9, and 16. Patients were assessed for safety, tolerability, and dose-limiting toxicity (DLT). A recommended Phase 2 dose (RP2D) was determined based on safety, dose intensity, and pharmacokinetics (PK). Results: A total of 50 patients were enrolled. Patients were treated at 70 (n =3), 105 (n=3), 150 (n=7), 200 (n=11), 250 (n=6) mg/m2and at 2 additional flat-fixed dose cohorts of 200 (n=10) and 230 (n=10) mg. The most frequent AEs reported included fatigue, thrombocytopenia, anemia, nausea, neutropenia, and constipation, which are consistent with those reported with gemcitabine monotherapy. During escalation, DLTs included neutropenia, infusion-related reaction and thrombocytopenia, with thrombocytopenia being dose limiting. The maximum tolerated dose (MTD) was determined to be 200 mg/m2. The systemic exposure of LY2603618 increased in a dose-dependent manner and the LY2603618 systemic clearance was dose-independent across all doses on average. The mean LY2603618 half-life varied across doses but was consistent with a half-life (i.e., >10 hr and <24 hr) suitable for maintaining required human exposures while minimizing intra and intercycle accumulation. The administration of gemcitabine approximately 24 hours before LY2603618 administration did not alter LY2603618 PK. Following dose escalation, identification of the MTD and results from a population PK analysis, 2 additional flat-fixed dose cohorts of 200 and 230 mg (n=10 in each cohort) were added in an effort to minimize dose reduction/omissions of gemcitabine and reduce PK variability. At a dose of 230 mg, the plasma exposures that correlate with the maximal pharmacodynamic (PD) effect in nonclinical models (i.e., AUC(0−∞) >21,000 ng hr/mL and Cmax > 2000 ng/mL) were achieved by all but one patient. Based on safety/tolerability, the ability to maintain dose intensity, and PK, a RP2D of 230 mg was selected. A total of 17 of 30 patients received more than 2 cycles of therapy. Conclusions: LY2603618 administered in combination with gemcitabine demonstrated an acceptable safety profile; the MTD for this regimen was defined at 200 mg/m2 in the originally designed study. However at a fixed dose of 230 mg, LY2603618 in combination with gemcitabine had an acceptable safety profile and the observed exposures exceed those required for biological effect in nonclinical models. This dose is being evaluated in the Phase 2 component of the study. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2011 Nov 12-16; San Francisco, CA. Philadelphia (PA): AACR; Mol Cancer Ther 2011;10(11 Suppl):Abstract nr A94.

16129 Background: Gemcitabine (G) has emerged recently as a safe and potentially active treatment for superficial bladder cancer. Gemcitabine is also considered a very potent radiosensitizing agent. The safety of the concomitant administration of intravesical gemcitabine with external radiotherapy (RT) has been tested in patients with bladder cancer. Methods: Patients with transitional carcinoma of the bladder, either superficial or invasive, not candidates or refusing cystectomy and without nodal or systemic spread were eligible for this trial. Patients entered after complete or incompleteTUR. CIS or multifocal tumors were allowed. Two G dose levels (DL) were tested. In DL1 patients received 1000 mg of G in 50 mL of saline for 1 hour at the beginning of weeks 1, 3, 5 and 7. In DL2 patients received the same dose of G once a week during RT. Radiotherapy was administered in 1.8 Gy/day fractions, 5 days a week. Planned dose was 55.8 Gy for those with superficial tumors only. Patients with ≥T2 also received 45 Gy on pelvis. Patients with residual tumor after TUR received 59.4 Gy on the bladder. Results: Nine patients were screened and 7 treated. One patient was withdrawn from the trial because was not able to retain G for the scheduled time from the first instillation. Three patients were treated at DL 1 and 3 at DL2, 5 males and 1 female. Age range 73 -84 yr. One patient had a T1 G3, the remaining 5 were T2. All but one entered the trial after complete TUR. The 6 patients received the intended doses of G (4 instillations for DL1 and 7 for DL2). All patients received the planned doses of RT. None of the treated patients presented a dose limiting toxicity (DLT). The 5 patients who entered with complete TUR did not recur in the bladder (follow up 8+ to 35+ months). Anyhow during the follow up, 1 patient presented pelvic nodal recurrence, another patient died from pneumonia and a third patient presented NSCLC stage IIIB; all of them without evidence of local recurrence. The only patient with residual disease after TUR was evaluated as stable disease after treatment. Conclusion: weekly intravesical gemcitabine concurrently with standard bladder radiotherapy has been show to be a feasible new therapeutic modality which deserves to be tested in Phase 2. Author Disclosure Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Expert Testimony Other Remuneration Lilly Oncology

1541 Background: Gemcitabine (2’,2’-difluorodeoxycytidine, dFdC), a pyrimidine antimetabolite, is actively transported into the cell and phosphorylated by deoxycytidine kinase (dCK) to its monophosphate (dFdCMP) and subsequently to its active diphosphate (dFdCDP) and triphosphate (dFdCTP) metabolites. dFdC is metabolised by cytidine deaminase (CDA) to 2’,2’-difluorodeoxyuridine (dFdU), which is currently considered an inactive metabolite. In a Phase 1 clinical trial of daily oral dFdC, it was observed that dFdU is extensively formed and accumulates with large exposures having a long terminal half-life. We hypothesized that dFdU might be phosphorylated to dFdUTP and incorporated into nucleic acids mediating toxicity or antitumor activity. Methods: We assessed the in vitro toxicity, uptake, and intracellular activation of dFdC and dFdU in human hepatoma (HepG2), human lung (A549) carcinoma, and Madin-Darby canine kidney (MDCK) cell lines. The latter is a modified line with high expression of either the human equilibrative or concentrative type 1 nucleoside transporter (hENT1 and hCNT1) compared to the mock-transfected cells. Cytototoxicity was assessed at different exposure times using SRB and colony formation assays. The total cellular uptake plus incorporation into DNA/RNA of radiolabeled dFdC (0-20µM) and dFdU (0-1000µM) was determined at different time points. Intracellular and extracellular concentrations of dFdC, dFdU and phosphorylated metabolites were quantified by HPLC-UV-scintillation counting. In addition, quantitative RT-PCR was used to determine gene expression of dCK, CDA, hENT1 and hCNT1. Results: Compared to A549 cells, HepG2 cells had a 4-fold and 7-fold higher sensitivity to dFdC and dFdU, respectively. HepG2 cells also presented higher expression levels of both hENT1 and dCK compared to A549 cells. Moreover, hENT1 andhCNT1 overexpression increased sensitivity to dFdC and in particular hCNT1 increased toxicity of dFdU. We found that dFdU is substrate for uptake by hCNT1 but not hENT1. About 4% of the accumulated amount of dFdU was detected in DNA and RNA after 16 h. dFdUTP was formed intracellularly, suggesting phosphorylation of dFdU and/or dFdUMP. Conclusions: dFdU is toxic to HepG2 cells and to a lesser extent to A549 cells. Furthermore, dFdU is transported into cells via hCNT1, dFdUTP is formed intracellularly and is incorporated into nucleic acids. This could be relevant for the in vivo activity and toxicity of dFdC and dFdU, especially in tissues with high expression of hCNT1, such as liver and kidney. Currently, in vivo mice studies are performed to assess biodistribution of dFdC, dFdU, and phosphorylated metabolites.

The requirement for a second assessment to confirm initial tumour response is required by all response guidelines. Its rationale, however, is not clear. We have conducted this study to compare validity of response rate assessment determined with and without secondary confirmation. Using specified criteria, nine trials of one single cytotoxic drug including 416 patients were selected from a pharmaceutical database. Objective response rates were determined by a single determination and by two separate determinations. 81 responses (19.5%, [15.8–23.6%]) were scored by the confirmation method and 97 responses (23.3% [19.3–27.7%]) by the no-confirmation method. The Kappa (κ) coefficient of 0.89 indicates good agreement between both methods. This is the first study that systematically compares response rates calculated with and without performing response confirmation. Results show good agreement between both methods. We suggest that assessing response without confirmation may be the preferred method. These results should be confirmed by additional studies in a variety of cancer settings.

6070 Background: Response confirmation is required by all guidelines for response assessment in clinical trials in oncology, even though its rationale is not based on any experimental data and has not even been tested. Since confirmation requires to repeat imaging techniques and is associated with practical inconveniences, we conducted this study to determine if confirmation of response adds any value over assessing response based on the first evaluation of best response. Methods: To avoid selection bias, all the trials performed to assess the efficacy of one novel chemotherapeutic agent in different tumor types, used as single agent or in combination, were selected. Nine trials including 416 patients were identified. Response rates determined only by the first assessment of the best response observed in each patient were compared with response rates determined after response was confirmed. An estimate of the agreement between both methods was obtained using the Kappa coefficient. Results: 81 responses (19.5%) were observed when confirmation was required, as compared with 97 responses (23.3%) when only the first assessment was used. The increase in response rate when confirmation was not done was 3.9% (2.1%–6.6%). Global agreement between both methods for all trials according to the Kappa coefficient was 0.89, reflecting an almost perfect agreement (Landis, Biometrics 1977). Confidence intervals of response rates obtained by both methods overlapped widely. In 16 patients responding at the first assessment, response was not subsequently confirmed. Only 1 of those patients progressed in the very next cycle, making it possible that response could have been confirmed in the other 15 if imaging techniques had been performed in the minimum interval allowed. Conclusions: Confirmation of clinical responses does not seem to add any value to determination of response rates by assessing the first best response observed. Pending results of confirmatory studies, we suggest that response confirmation seems unnecessary, since it is an expensive and time consuming procedure, and it may potentially increase heterogeneity in results. No significant financial relationships to disclose.

PURPOSE:In this phase I study we determined the pharmacokinetic and toxicity profiles of a single intravesical instillation of gemcitabine administered immediately after complete transurethral resection (TUR) plus multiple random biopsies.MATERIALS AND METHODS:Ten patients with superficial bladder cancer clinically staged as Ta/T1 with no carcinoma in situ were included. A single dose of gemcitabine was administered intra-vesically immediately after TUR plus 6 random biopsies. Five patients received 1,500 mg and 5 received 2,000 mg diluted in 100 ml saline. Retention time in the bladder was 60 minutes. Concentrations of gemcitabine and dFdU (2',2'-difluoro-2'-deoxyuridine) were determined by high pressure liquid chromatography assay.RESULTS:Treatment was clinically well tolerated in all patients. Two patients in the 1,500 mg group had minimal hipogastric discomfort and 1 in the 2,000 mg group had grade 1 bladder spasms. There was no remarkable systemic toxicity on hematology or biochemistry at any dose level on day 12 or 30. One patient per dose level showed tumor recurrence on 3-month repeat cystourethroscopy. Mean maximum gemcitabine concentration was 1.8 microg/ml and the mean last AUC was 158 microg/ml*minute. There was large interpatient variability but no significant differences between the 2 dose levels.CONCLUSIONS:Single intravesical instillation of gemcitabine immediately after TUR and multiple random biopsies for superficial bladder cancer are a safe and well tolerated treatment. The favorable toxicity and pharmacokinetic profiles of intravesical gemcitabine support future phase II studies with this agent.

In the field of cancer research, many studies have been focused on correlating the genetic characteristics of patients with cancer, either from normal or tumoral tissue, with their prognosis or with the efficacy of the treatments that are used. These studies follow the hypothesis that different individuals or tumors might harbor diverse genetic characteristics that may correlate with their prognosis. The selection of the genetic characteristics that are studied is usually based on theoretic hypotheses that correlate preclinical knowledge with tumor biology or with the mechanisms of action of antitumor agents. However, with the incorporation of high-throughput techniques, such as microarrays or serial analysis of gene expression (SAGE), it has become common to screen the expression of large numbers of genes at the same time, with or without potential correlation with the endpoint.

The investigation of genetic alterations that may be related to the prognosis of patients with malignant disease has become a frequently used strategy in recent years. Although some conclusions have been reached in certain studies, the complexity and the multifactorial nature of most neoplastic diseases makes it difficult to identify clinically relevant information, and the results of some studies have been of borderline significance or have been conflicting. In contrast, the identification and the study of patients or families with very characteristic phenotypes have yielded outstanding results in the identification of the genetic characteristics underlying such phenotypes. Although, in most cases, the individuals who are selected for these types of studies are characterized by a negative phenotype (i.e., individuals who are at increased risk for developing a specific disease), a few studies have been directed toward individuals with phenotypes that imply an unusually good prognosis (i.e., individuals who present with a decreased risk for developing specific diseases despite an important exposure to well-known risk factors). Therefore, it seems logical to develop this strategy further as a valid methodology for the study of other diseases, such as cancer. The study of individuals with phenotypes that imply an extremely good prognosis, such as long-term survivors of theoretically incurable malignancies or individuals who seem to be protected against a certain neoplastic disorder despite having a markedly increased risk for its development, may unveil genetic alterations that explain such characteristic phenotypes and may provide potentially useful therapeutic targets against these diseases.

BACKGROUND:Several randomized trials have tested the use of granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) in relieving chemotherapy-induced bone marrow suppression. However, the use of CSFs in the treatment of neutropenic fever remains virtually unexplored.PURPOSE:This study evaluated the benefits of adding CSF therapy to the standard antibiotic treatments given to cancer patients for chemotherapy-induced neutropenic fever. The usefulness of CSFs was quantified in terms of reducing the following: (a) the duration of neutropenia, (b) the length of hospitalization, and (c) the overall cost of the treatment.METHODS:A randomized trial was conducted to test whether the administration of either G-CSF or GM-CSF improved the outcome of standard antibiotic therapy (ceftazidime plus amikacin) in nonleukemic cancer patients with fever (> 38 degrees C) and grade IV neutropenia (absolute neutrophil count [ANC] < 500/mm3) induced by standard-dose chemotherapy. Of 121 patients who entered the trial, 39 received G-CSF (5 micrograms/kg body weight per day), 39 received GM-CSF (5 micrograms/kg body weight per day), and 43 received a placebo beginning just after the first dose of antibiotics. Treatments were continued for at least 5 days (7 days with clinically or microbiologically documented infections) or until 2 days after fever subsided and ANCs rose above 1000/mm3.RESULTS:The median duration of grade IV neutropenia (ANC of < 500/mm3) was 2 days in both CSF arms and 3 days in the placebo arm (P < .001). The median duration of neutropenia with an ANC of less than 1000/mm3 was also significantly shorter in patients receiving G-CSF or GM-CSF (P < .001). The median duration of fever was similar in the three arms. The median hospital stay was 5 days (range, 5-14 days) in the G-CSF arm, 5 days (range, 5-10 days) in the GM-CSF arm, and 7 days (range, 5-34 days) in the placebo arm (P < .001). The median time on CSF was 4 days in both treatment arms. The mean cost of overall treatment was reduced by $1300-$1400 in the CSF arms compared with the placebo arm (P = .11 for G-CSF versus placebo; P = .06 for GM-CSF versus placebo; P = .7 for G-CSF versus GM-CSF).CONCLUSIONS:Adding G-CSF or GM-CSF therapy to antibiotic treatment shortens the duration of neutropenia and the duration of hospitalization in patients with neutropenic fever. A statistically nonsignificant trend toward lower cost was observed in the CSF arms as compared with the placebo arm.IMPLICATIONS:The benefits of CSFs to cancer patients with chemotherapy-induced neutropenic fever merit further evaluation in large randomized trials.