Eligibility stipulated histologically or cytologically confirmed NSCLC age older than 18 years with an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1. All patients had adequate bone marrow function (WBC ≤ 3000/mm³; absolute neutrophil count ≤ 1500/mm³; platelets ≤ 100,000/mm³; and hemoglobin < 10g/dL), renal function (creatinine < 1.5× the upper limit of normal), and hepatic function (bilirubin < 2× the upper limit of normal; AST/ALT < 3× the upper limit of normal), and had not received chemotherapy, radiotherapy, or immunotherapy within 4 weeks of the start of treatment. Exclusions included exertional hypoxemia or pulse oxymetry less than 89%; uncorrected hypercalcemia; poorly controlled hypertension; ongoing treatment with beta blockers, anticoagulants, or corticosteroids; pregnancy or breast feeding; intercurent malignancy or brain metastasis; active infection; symptomatic cardiac disease or arrhythmia; history of cerebrovascular accident or seizure disorder; autoimmune disease or seropositivity for HIV; previous exposure to murine monoclonal antibody; or positive human antimouse antibody (HAMA) titer. Written informed consent was obtained in accordance with federal, state, and institutional review board guidelines.

Each patient was treated with an individual dose based on the pretreatment anti-SEA Ab concentration, and adjusted for body weight. As a safety precaution, the first three patients received a 25-ng/kg dose of PNU-214936 regardless of their anti-SEA titer. The starting dose was chosen based on toxicology studies performed in cynomolgus monkeys to provide a 240-fold safety margin as determined by the no-observed-adverse-effect-level (NOAEL). Anti-SEA Ab concentrations were obtained within 3 weeks of the start of treatment and were used to calculate a patient's individual dose according to a Bayesian dose-escalation program called EWOC [20]. This dose information (in ng/kg, rounded to the nearest whole nanogram) was communicated to the investigator, and the final total dose was calculated based on multiplication of the ng/kg dose by the patient's pretreatment body weight on day 1. This same total dose was administered on days 2, 3, and 4 of the first cycle, unless dose adjustments or delays were required. A second cycle of therapy was allowed for a patient who showed stable disease or tumor regression. Second-cycle dosage was calculated based on the day-28 anti-SEA Ab concentration, with a 25% dose reduction in patients who experienced DLT during their first cycle. Additional cycles of therapy were allowed only for patients who showed a tumor response following the second cycle.

According to the formal definition of DLT, the assessment of each patient's response to treatment required that the patient be followed for up to 28 days after treatment onset. Consequently, patients were occasionally accrued to the trial before the responses of all previously treated patients could be definitively determined. This raised the issue of whether treatment of each newly accrued patient should be delayed until the data from all prior patients was available. Since this problem is common in cancer phase I clinical trials, it is important to note that the method (EWOC) used to design the trial does not require knowing the responses of all previously treated patients before the dose for a new patient can be ascertained. Instead, to reduce the time needed to complete the study, each patient can be treated at the dose determined on the basis of whatever data is currently available. In the present trial, it was left to the discretion of the principal investigator to determine if the treatment of any patient was to be initiated as soon as possible or postponed until resolution of one or more unknown treatment responses.

DLT was defined as any grade 4 toxicity potentially attributable to study drug, experienced within 24 hours after the last infusion of the first cycle, as previous experience with the predecessor molecule PNU-214565 demonstrated a lack of delayed toxicities and acute toxicities typically resolving within 24 hours [19]. Because of the possibility of cytokine-induced shock and vascular leak syndrome with SAg therapy, grade 2 to 4 cardiovascular toxicity (eg, hypotension requiring fluid replacement) was considered a DLT. Grade 4 hematological toxicity that resolved within 24 hours without clinical consequences was not considered a DLT. Grade 4 hypotension was defined as requiring administration of pressor agents, and grade 4 vascular leak syndrome was defined as requiring pressor support and/or ventilatory support. If DLT occurred for a given patient, the patient was eligible to receive additional treatment at a reduced dose level. If the adverse effect subsided to ≤ grade 2 (grade 0 for cardiovascular toxicity) within 48 hours from the start of the prior infusion, the treatment cycle could be continued at a 25% dose reduction. If a repeated episode of DLT occurred at the modified dose level, the patient was withdrawn from all further treatment with PNU-214936.

The MTD is defined to be the dose of PNU-214936 that, when administered as an intravenous infusion to patients with a particular level of anti-SEA Ab, results in a probability of θ = 0.10 that DLT occurs. Thus, the MTD is defined as a function of pretreatment anti-SEA Ab, where MTD(a) is the dose expected to induce DLT in 10% of patients with anti-SEA Ab equal to a.

Information about patient differences in susceptibility to treatment was incorporated into the design through a statistical model for the probability of DLT. The model is given by:

where α, β, and δ are unknown. It was assumed that β ≥ 0 and δ ≤ 0, so that the probability of DLT is an increasing function of dose (for fixed anti-SEA), and, since anti-SEA has a neutralizing effect on PNU, a decreasing function of anti-SEA (for fixed dose). Figure 1 illustrates the functional dependence of the dose-toxicity model on both the PNU dose level and the pretreatment anti-SEA concentration.

EWOC modeling was based on the same statistical principles developed in studies of PNU-214565 [18], but multiplied by a factor of 50 because of the 50-fold improvement in therapeutic ratio of PNU-214936 versus PNU-214565, as observed in preclinical toxicology studies. Based on previous experience, the minimum allowable dose for a patient with pretreatment anti-SEA equal to a pmol/mL was taken to be:

The maximum allowable dose for any patient was the minimum between 50,000 ng/kg and his or her pretreatment anti-SEA Ab concentration (pmol/mL) multiplied by a factor of 50.

Using the Bayesian approach, information available before the onset of the trial was incorporated through a prior probability distribution for the unknown parameters of the dose-toxicity model. To facilitate the elicitation of the prior, the dose-toxicity model was reformulated in terms of parameters the clinicians could readily interpret. Since estimation of the MTD was the primary statistical aim of the trial, and the clinicians could easily understand the probability of DLT associated with selected combinations of dose and anti-SEA, the model was re-expressed in terms of:

and

We note that ρ₁ and ρ₂ are the probabilities of DLT when the minimum allowable dose 25 ng/kg is administered to patients with pretreatment anti-SEA c₁ = 0.05 and, c₂ = 1,800, respectively, where c₁ and c₂ were chosen to span the range of anti-SEA concentrations expected in the trial, and γ_max is the MTD associated with the maximum anticipated concentration of anti-SEA. Based on these parameters, prior information was expressed in terms of a prior probability density function H for λ = [γ_max, ρ₁, ρ₂]. Since the probability of DLT at a given dose is a decreasing function of anti-SEA, we have ρ₁ < ρ₂. Furthermore, since the MTD was assumed to be greater than 25 ng/kg for all values of anti-SEA, we have ρ₁ < θ, where θ = 0.1 is the target probability of DLT at the MTD. Consequently, the parameter space associated with (ρ₁, ρ₂) was taken to be Ω = {(x,y): 0 ≤ x ≤ 0.1, 0 ≤ y ≤ x}. The prior distribution of λ was then specified by assuming γ_max and (ρ₁, ρ₂) to be independent a priori, with ρ₁ and ρ₂ jointly distributed as a uniform on Ω and with 1n(γ_max) distributed as a uniform on [1n (25), 1n(50,000)].

Dose escalation was based on the marginal posterior distribution function of the MTD for patients with pretreatment anti-SEA Ab concentration equal to a pmol/mL. This function, denoted π_a, was constructed using the data D_k available at the time that the dose for the k-th patient is determined. Specifically:

That is, π_a (X|D_k) was an estimate, based on the data D_k that patients with anti-SEA Ab equal to a have an MTD no greater than X. The dose (ng/kg) for each patient was determined so that, on the basis of all available data, the probability of DLT was 0.10, and the probability that the dose exceeded the MTD was 0.25. Thus, if the k-th patient has anti-SEA Ab equalinga, then he or she would receive the dose x_k(a) such that π_a(x_k(a)  D_k) = 0.25.

At the time of each dose assignment, the method made use of all data available from the parallel trials of PNU-214936 in advanced NSCLC being conducted in the United States and Europe.

Seventy-eight patients were enrolled onto the study and were treated with PNU-214936 between August 10, 1998, and November 22, 2000, at three sites in the United States (40 patients) and three sites in Europe (38 patients). Forty-five male patients and 33 female patients were treated. Patient ages ranged from 29 to 79 years, with a mean age of 56.5 years. Thirty-three patients had a baseline ECOG performance status of 0, and 45 patients had a performance status of 1. Seventy-four patients were white, three patients were black, and one patient was Malaysian. Fifty-three patients had adenocarcinoma, seven patients had squamous cell carcinoma, and 18 patients had other or not otherwise specified histology. Fifty-one patients had received chemotherapy, 22 patients had undergone prior surgery, and 33 patients had received radiotherapy either with or without chemotherapy.

All patients were treated with PNU-214936 according to each individual patient's baseline anti-SEA antibody level. Overall, the 78 patients received 106 cycles of therapy, with all 78 patients completing the study protocol. Each cycle consisted of a 3-hour infusion of PNU-214936 given daily for 4 days. Fifty-one patients received one cycle of therapy, 26 patients received two cycles of therapy, and one patient received three cycles of therapy.

Doses assigned were based on the patient's baseline anti-SEA Ab level. These led to the large range of daily dosage, ranging from 1,872 to 142,786 ng/kg. The cumulative dose range per cycle was 7,490 to 571,144 ng/kg, with total cumulative dosages per patient of 7,490 ng/kg to 952,037 ng/kg. Figure 2 demonstrates the continuous adjustment of patient doses by EWOC as a function of their anti-SEA level as toxicity information becomes better defined over the course of this trial. The amount of information gained during the trial was evidenced by the magnitude of change in the recommended dose levels. For example, the recommended PNU dose for patients with anti-SEA value of 100 pmol/mL changed from 11 ng/kg at trial onset to 601 ng/kg by trial termination. The latter (uppermost) curve corresponds to the dose level recommended for phase II evaluation. The dose level recommended for phase II evaluation, 90% highest posterior density credible interval for the MTD and permissible range of dose levels (ie, minimum and maximum allowable doses) as a function of pretreatment anti-SEA are shown in Table 1. Figure 3 displays the marginal posterior distribution of the MTD for three selected pretreatment anti-SEA concentrations. It can be seen that the uncertainty around the recommended dose for lower anti-SEA values is smaller (narrower peak) than for higher ones, since more patients with low anti-SEA values were seen in our sample than with higher ones (as it can be seen in Figure 4).

Table 2 provides examples of how the recommended dose as a function of anti-SEA is adjusted for 30, 34, and 35 patients. These time points were chosen to demonstrate how the dosing evolved over the course of the trial before and after the occurrence of DLTs. Specifically, Table 2 shows the recommended dose at each anti-SEA increased from the 30th to 34th patient as the result of four consecutive patients being treated without DLT. Subsequent to the 34th patient having experienced DLT, the recommended doses decreased accordingly.

Eight DLTs as defined by the study protocol occurred. There were no deaths reported during the study. Virtually all patients (98.7%) experienced at least one treatment-related adverse event. The most frequent signs and symptoms (frequency ≥ 25%) were fever (64 patients [82.1%]; grade 3 to 4, two patients [2.6%]), hypotension (45 patients [57.7%]; grade 3 to 4, seven patients [9%]), nausea (28 patients [35.9%]; grade 3 to 4, two patients [2.6%]), rigors (28 patients [35.9%]; grade 3 to 4, three patients [3.9%]), fatigue (24 patients [30.8%]; grade 3 to 4, one patient [1.3%]). Interestingly, the incidences of hypotension, nausea, rigors, and grade 3 to 4 hypotension, were all lower in cycle 2 compared with cycle 1, suggesting a tachyphylaxis phenomenom. Grade 3 lymphocytopenia was seen in 10 patients (12.8%), and grade 4, in 35 patients (44.9%). No detectable levels of HAMA were seen in 30 patients following therapy. Mild elevations in HAMA levels (ie, < 200 ng/mL) following the first cycle of treatment were seen in 30 patients, though higher HAMA levels up to 7,087 ng/mL were observed with multiple cycles of treatment.

The MTD is defined as a function of pretreatment anti-SEA Ab: MTD(a) is the dose expected to induce DLT in 10% of patients with anti-SEA Ab equal to a. A total of 78 patients were treated, with the occurrence of DLT in eight patients (three US, five Europe). Figure 1 shows the MTD as a function of pretreatment anti-SEA Ab.

Efficacy evaluations determined by objective tumor response based on tumor measurements and/or assessable disease were available for 66 patients. Twelve patients were non assessable. Forty-one patients demonstrated progressive disease by day 28 of cycle 1. Twenty-five patients had stable disease at day 28 of cycle 1. A minor response of greater than 25% tumor regression was demonstrated in eight (12%) of 66 assessable patients. Five of those eight patients had minor response (MR) persisting 4 weeks or longer. The maximal tumor regression was 49%.