Targeting tumour vasculature as a treatment for cancer is the subject of much research, with most effort currently directed at angiogenesis inhibitors. Yet therapies that interrupt existing tumour vasculature and result in haemorrhagic necrosis of tumours have been investigated for over a century and were mostly associated with bacterial infections or their toxins (Chaplin and Dougherty, 1999). William Coley, who pursued the method avidly from 1891 (Council on Pharmacy and Chemistry, 1934), achieved significant cure rates, especially in sarcoma and lymphoma, using various bacterial toxins both for advanced disease and as adjuvant therapy (Coley-Nauts et al, 1953). Coley's toxins were abandoned in favour of radiotherapy, but more recent clinical studies with bacterial endotoxins have shown modest antitumour activity, although systemic toxicity remains dose-limiting (Otto et al, 1996; DeVore et al, 1997). In 1947, investigators at the National Cancer Institute observed that a toxin from the bacterium Serratia marcescens acutely and irreversibly reduced blood flow to sarcomas in mice, while blood flow to muscle recovered within 18 h (Algire et al, 1947). The induction of tumour necrosis factor (TNF) by bacterial toxins was later shown to be the cause of haemorrhagic necrosis in tumours (Carswell et al, 1975; Old, 1985). Clinical trials of recombinant human TNF were unable to deliver therapeutic doses systemically because of toxicity (Spriggs and Yates, 1992), but TNF is highly effective in conjunction with chemotherapy in isolated limb perfusion for melanoma and sarcoma, where it has been shown to have selective tumour vascular effects (Lejeune et al, 1998; Ruegg et al, 1998; Eggermont and ten Hagen, 2001). Nonbacterial, tumour vascular-targeting agents could have advantages over bacterial products in terms of toxicity and pharmacology. One such agent, flavone acetic acid (FAA), showed remarkable anticancer activity in murine models (O'Dwyer et al, 1987) but its activity was minimal in clinical trials, although these were conducted when it was thought to be directly cytotoxic and its indirect mechanisms of antitumour action were poorly understood (Kerr and Kaye, 1989; Bibby, 1991). Further preclinical studies at the Auckland Cancer Society Research Centre (ACSRC) showed that FAA induced TNF production, acute tumour vascular collapse, haemorrhagic necrosis of tumours and enhanced activity of immune effector cells (Ching and Baguley, 1987; Smith et al, 1987; Finlay et al, 1988; Zwi et al, 1989, 1992) (Zwi et al, 1990 ID: 286). Few appropriate pharmacodynamic (PD) studies were performed in the clinical trials of FAA, so it is not known whether it has significant antivascular or immunological activity in humans. However, in vitro studies suggested a species difference in antitumour activity, because FAA induced TNF in murine, but not human, peripheral blood mononuclear cells (Ching et al, 1994; Philpott et al, 1997). 5,6-Dimethylxanthenone-4-acetic acid (DMXAA), a new analogue of FAA developed in the ACSRC, showed greater activity and 12-fold higher dose-potency than FAA in murine tumour models (Rewcastle et al, 1991) and appeared to overcome the species difference in in vitro TNF production (Ching et al, 1994; Philpott et al, 1997). Dose-limiting toxicity in mice was consistent with hypotension, and was not strictly related to TNF production (Philpott et al, 1995). While DMXAA proceeded into clinical development because of this favourable profile, further studies in animal tumour models have demonstrated synergistic interactions between DMXAA with radiotherapy (Wilson et al, 1998), chemotherapy (particularly taxanes) (Pruijn et al, 1997; Horsman et al, 1999; Wilson and Baguley 2000; Siim et al, 2002), bioreductive cytotoxic drugs (Cliffe et al, 1994; Lash et al, 1998), radioimmunotherapy (Pedley et al, 1996), antibody-directed enzyme prodrug therapy (ADEPT) (Pedley et al, 1999), thalidomide (Cao et al, 1999) and immunotherapy (Kanwar et al, 2001). Cancer Research UK selected DMXAA for two parallel phase I trials with different schedules, one in the UK using a weekly administration schedule (at Mt Vernon Hospital, Northwood, and Bradford Royal Infirmary, Bradford) and the other in Auckland, NZ, dosing every 3 weeks. The objectives of the NZ study were to determine the toxicity of DMXAA, its maximum tolerated dose (MTD), pharmacokinetics (PK), selected PD end points and, as a secondary objective, antitumour efficacy. A protocol amendment permitted evaluation of changes in tumour blood flow by dynamic magnetic resonance imaging (MRI) and the findings (including both UK and NZ trial patients) have recently been published (Galbraith et al, 2002). This report focuses on clinical aspects of the NZ trial; PD and PK aspects, due to their complexity, will be reported separately. MATERIALS AND METHODS Patients Eligibility criteria included: patients ⩾18 years with histologically or cytologically proven cancer refractory or not amenable to conventional therapy; WHO performance status of 0–2; life expectancy >3 months; haemoglobin ⩾9 g dl−1, WBC ⩾3 × 109 l−1, platelets ⩾100 × 109 l−1; creatinine ⩽130 μmol l−1; bilirubin within normal limits, ALT and alkaline phosphatase 450 ms in males, >470 ms in females) after treatment with doses of DMXAA of ⩾3100 mg m−2 (Committee for Proprietary Medicinal Products, 1997). The maximal QTc prolongation occurred during or within 15 min of completion of the drug infusion with a return to baseline generally over 4–6 h (Figure 2 Figure 2 Change in corrected QT interval after DMXAA administration at 3700 mg m−2. ). QTc was prolonged beyond 500 ms in four patients with baseline values of 430, 460, 461 and 501 ms and in the first three patients, this lasted for less than 1 h. No clear dose–response relationship was observed and no ventricular tachyarrhythmia was seen. Dyspnoea possibly related to DMXAA was documented in three patients treated at doses ⩽1100 mg m−2, but each patient had another more likely explanation. At higher doses, acute dyspnoea at rest occurred in four patients immediately following DMXAA infusion. In three of these patients (treated at 2600–3700 mg m−2), the dyspnoea after DMXAA was minimal, brief (5–25 min) and clinically insignificant. The fourth patient, with moderately severe chronic obstructive respiratory disease, became very breathless and anxious 15 min after her first course at 4900 mg m−2. Clinically, she had basal pulmonary inspiratory crackles but no bronchospasm. Oxygen saturation was normal on breathing air and the symptoms resolved in about 75 min. On her second course, DMXAA was reduced to 3700 mg m−2 (she is thus represented at two doses in Table 3) and she again developed dyspnoea, lasting for about an hour. She had basal inspiratory crackles and a third heart sound (consistent with left ventricular failure) and these signs resolved within 2 h. Therapeutic response A total of 60 patients were evaluable for response. One partial response was seen in a patient with metastatic squamous carcinoma of cervix treated at 1100 mg m−2 and previously treated with bleomycin, etoposide and cisplatin chemotherapy, then carboplatin. The response was unconfirmed because two small neck nodes increased in size transiently after the third course then subsequently regressed again and overall tumour response was maintained for eight courses (Figure 3 Figure 3 Tumour response in a patient with metastatic squamous carcinoma of cervix treated with DMXAA 1100 mg m−2. Tumour size was calculated as the sum of the products of bidimensional measurements of three clinically measurable metastatic neck nodes. ). No change was the best response in 22 patients (35%) at dose levels from 6 to 3700 mg m−2 and the duration exceeded 12 weeks in five patients (8%). DISCUSSION This phase I trial of DMXAA has demonstrated clinical antitumour activity and reduction in tumour blood flow at well-tolerated doses (Galbraith et al, 2002). The MTD was established as 3700 mg m−2, and higher doses produced unusual, transient toxicities including confusion, slurred speech, tremor, restlessness and visual disturbance, as well as transient autonomic changes such as sweating, tremor, changes in blood pressure and heart rate. The results of the UK trial of DMXAA are published separately (Rustin et al, submitted to Br J Cancer), but there was a high degree of concordance with our results including toxicities, PK, PD and clinical activity (an unconfirmed partial response was also seen). Clinical trials of FAA, which is structurally related to DMXAA, showed differences between the two drugs. While the clinical toxicities of DMXAA and FAA showed some similarities (warmth, flushing, sweating, fatigue, myalgia, nausea, vomiting and visual disturbance), the DLTs of FAA were remarkably different and included hypotension, diarrhoea, flushing, asthenia and fatigue (Kerr et al, 1987; Weiss et al, 1988; Kaye et al, 1990; Havlin et al, 1991; Olver et al, 1992). Prolongation of bleeding time was described (Rubin et al, 1987) and one patient presented with haemorrhage due to immune thrombocytopenia after five doses of FAA (Davis et al, 1988). Bleeding time was not measured in this DMXAA trial, but no significant disturbance of INR and APTT was seen, nor was any significant thrombocytopenia observed. A number of other agents, including serotonin, some tubulin-binding agents and arsenic trioxide, selectively inhibit tumour blood flow (Stücker et al, 1992; Ching et al, 1999; Lew et al, 1999). Combretastatin A-4 phosphate (CA4P), a tubulin-binding agent, has antivascular activity at doses well below the MTD in both preclinical and early clinical studies (Rustin et al, 1999), comparable to those reported with DMXAA (Lash et al, 1998; Galbraith et al, 2002). However, the blood flow-modifying effect of CA4P is not entirely tumour-selective (Griggs et al, 2001), and it appears to be reversible except at high doses (Anderson et al, 2000; Murata et al, 2001). Some CA4P toxicities resembled those of DMXAA (including flushing, nausea, vomiting, tumour pain and QTc prolongation), but others (cardiac ischaemia and cerebellar ataxia) were notably different (Galbraith et al, 2001; Dowlati et al, 2002). Some toxicities observed with DMXAA resemble the ‘serotonin syndrome’, attributed to high levels of serotonin in the central nervous system (CNS). This syndrome most commonly occurs when two drugs are taken, which can each increase CNS serotonin and includes alterations in cognition, behaviour, autonomic nervous system function and neuromuscular activity (Sporer, 1995). Supportive evidence comes from two other aspects of this trial: firstly, a patient took a monoamine oxidase inhibitor prior to treatment with DMXAA at 1650 mg m−2 and developed clinical features of the serotonin syndrome; and secondly an acute increase in plasma prolactin levels seen in many patients treated with DMXAA at ⩾2000 mg m−2 (unpublished results). This occurs following production of serotonin in the CNS (Van de Kar et al, 1996). If subsequent studies provide further evidence that DMXAA increases CNS serotonin release, it would be prudent to avoid administering DMXAA to patients who are receiving other drugs that increase CNS serotonin levels (Brown et al, 1996). The mechanism underlying the acute release of CNS serotoninis not known, but serotonin release in plasma has been observed in this study (Kestell et al, 2001), and is a feature of the antivascular activity of this drug in preclinical models (Baguley et al, 1997). The visual toxicities of DMXAA are not a feature of the serotonin syndrome, but the blurring, colour disturbance and photophobia have similarities to those reported with sildenafil (Viagra®, Pfizer, New York, NY, USA) (Marmor and Kessler, 1999). The latter's visual toxicities are thought to be due to inhibition of phosphodiesterase type 6 (PDE6), which exists exclusively in the retina and is responsible for modulating the transduction cascade of the photoreceptor response to light. Recent data have shown that DMXAA inhibited PDE6 in vitro at pharmacologically relevant concentrations (e.g. 50 μ M) (personal communication, L Kelland, Antisoma plc). It is reassuring in this regard that no long-term retinal sequela of sildenafil administration is known, including data from retinal histologic studies in dogs dosed with 65 times the maximum recommended human dose daily for 12 months (Wallis et al, 1998; Marmor and Kessler, 1999). The significance of the observed QTc prolongation in these preliminary data is uncertain given its brevity, its possible relation to autonomic changes caused by the infusion and the population of patients potentially being treated with this agent. Moreover, the use of ad hoc heart rate correction formulae (such as Bazett's) may bias the result (Aytemir et al, 1999). Therefore the QTc results in this trial must be regarded as indicative only. However, given the concern that QT interval prolongation may predispose to ventricular tachycardia (Committee for Proprietary Medicinal Products, 1997), it will be important to determine the dose–response relationship of QTc prolongation before deciding on the dose of DMXAA for phase II and combination studies. In conclusion, DMXAA is well tolerated over a wide dose range and has clinical and biological features distinct from those of other antivascular drugs in clinical development. Further clinical trials with this agent are clearly warranted, particularly in combination with other treatment modalities where synergistic interactions are observed in animal tumour models.