NDA # : 21-236

Applicant : Matrix Pharmaceutical, Inc.

Name of the Drug: IntraDose (cisplatin/epinephrine) Injectable Gel

Indication : Squamous Cell Carcinoma of the Head and Neck

Medical Reviewer : Gregory Frykman, M.D.

Statistical Reviewers : Jasmine Choi, M.S., Rajeshwari Sridhara, Ph.D.

________________________________________________________________________

I.C.H. E-9 Guidelines: Section 2.1.2, Confirmatory Trial:

I.C.H. E-9 Guidelines: Section 2.2.2, Primary and Secondary Variables:

I.C.H. E-9 Guidelines: Section 2.2.3, Composite Variables:

I.C.H. E-9 Guidelines: Section 2.2.5, Multiple Primary Variables:

• There are two co-primary efficacy endpoints in each of the two randomized studies (Studies 414 and 514), namely, objective tumor response rate and patient benefit. The understanding between the agency and sponsor was that the trials will demonstrate effect on both of the primary variables, thus not requiring adjustment of type I error for multiplicity (I.C.H. E-9 Guidelines, section 2.2.5). The sponsor claims significant treatment (Intradose) effect with respect to objective tumor response rate. However, the sponsor's analyses demonstrate no significant treatment effect with respect to patient benefit.
• In addition to significant tumor response, the agency also required that significant correlation between objective tumor response and patient benefit should be established. There is no statistically significant association between objective tumor response and patient benefit in Study 414 (US study), where as there appears to be nominal significant association in Study 514 (Europe study). This apparent association is weak due to large number of patients classified both as non-responders and non-benefitters. Study 414 and Study 514 demonstrate that in these two studies it does not provide a good measure of predictability of benefit using tumor response. A preferred measure of association between patient benefit and tumor response is sensitivity (probability of benefit and response), and this was < 50% in each of the studies and in the combined study data.
• Patient benefit, one of the co-primary endpoints, is retrospectively defined as a decrease in symptom score by one point. Furthermore, the treatment goal differed from patient to patient.
• Palliative symptom benefit is assessed as equivalent to preventive benefit. Preventive benefit is not a measure of symptom benefit, but a measure of progression of disease.
• Blinded comparison of patient benefit between the treatment arms is questionable because the pattern of treatment conformity in the blinded phase differs between the active treatment and the placebo arms. This could potentially bias the benefit scores, particularly preventive goal scores.
• Dosing scheme of Intradose was modified after approximately 50% of the patients had been enrolled into the two studies. Because tumor response is a primary endpoint and the proposed treatment is a local treatment, there is a potential for dose-response relationship to play an important role in determining efficacy of the drug.

1. Background

Cisplatin/epinephrine (CDDP/epi or Intradose) injectable gel is proposed to be used for the local treatment of recurrent or refractory squamous cell carcinoma of the head and neck (SCC H&N) in patients who are considered not curable with surgery or radiotherapy. The intent of this injectable gel system is to place the product directly into the tumor tissue, providing high local concentrations of the active drug (cisplatin) that are maintained in the tumor (i.e., at or near the site of administration) over an extended time period. The function of the collagen is to provide a dense meshwork that gives the formulation its physical and drug retention properties. This viscous gel matrix entraps the cisplatin, providing a uniform suspension of the drug, and is believed to provide enhanced drug retention by inhibiting local interstitial fluid flow and concomitant drug clearance. The formulation also contains epinephrine, acting as an adjuvant to effect local vasoconstriction, further inhibiting cisplatin clearance from the administration site by restricting local blood flow, and providing additional enhancement of local drug retention. CDDP/epi gel is intended for intratumoral injection only. Solid tumors that can be seen, palpated, or visualized with established imaging techniques or accessed directly or by means of minimally invasive techniques, may be potentially treated.

2. Description of Trials

1. Study 39-92-P: Phase I, open-label, multi-center dose escalation/dose-confirmation study in patients with accessible tumors of any histology (45 patients studied in US).
2. Study 403-93-2: Phase II, open-label multicenter study in patients with recurrent or refractory accessible tumors of any histology except Kaposi's sarcoma and squamous cell carcinoma of the head and neck (67 patients studied in US).
3. Study 503-93-2: Phase II, open-label multicenter study in patients with recurrent or refractory accessible tumors of any histology except Kaposi's sarcoma and squamous cell carcinoma of the head and neck (59 patients studied in Europe and South Africa).
4. Study MP 414-94-2: Phase III, randomized, double-blind , placebo-controlled, multicenter study with an open-label extended follow-up phase for patients with squamous cell carcinoma of the head and neck (110 patients in US and Canada).
5. Study MP 514-94-2: Phase III, randomized, double-blind , placebo-controlled, multicenter study with an open-label extended follow-up phase for patients with squamous cell carcinoma of the head and neck (115 patients in Europe and Israel).

Studies 39-92-P, 403-93-2 and 503-93-2 are non-randomized, single arm open-label studies. Hence, this review will focus only on the two randomized studies, MP 414-94-2 and MP 514-94-2, and specifically on the efficacy aspect of these two studies.

2. Detailed Description of Studies MP 414-94-2 and MP 514-94-2

Study MP 414-94-2 was conducted in U.S. & Canada (44 study centers), and Study MP 515-94-2 was conducted in Europe & Israel (28 study centers). Both the studies MP 414-94-2 (here after referred as Study 414) and MP 514-94-2 (here after referred as Study 514) were designed identically as phase III, randomized, double-blind, multi-center, placebo-controlled studies to evaluate efficacy and safety of the treatment with CDDP/epi gel in patients with recurrent or refractory squamous cell carcinoma of head and neck. Patients were stratified according to baseline volume of the pre-specified Most Troublesome Tumor (MTT) or the primary target tumor (Stratum 1: MTT £ 5 cm³; Stratum 2: MTT >5 but £ 20 cm³) and then randomized in a 2:1 ratio to receive either CDDP/epi gel or the placebo. A stratum 3 (MTT > 20 cm³ ) was included to study safety in expanded use but not included as part of the pivotal analysis of safety and efficacy. Enrollment of patients to Stratum 3 was closed as of April 4, 1997 (Study period was from June 1995 to March 2000). Per protocol, prior to enrollment of the patient into the study, the investigator was required to identify the patient's most troublesome tumor (MTT) and one improvable primary treatment goal (palliative or preventative) (see Treatment Goal Questionnaire in Appendix 1) for that tumor. The patient could also identify a most troublesome tumor and may select one primary treatment goal (palliative), for this tumor. Patient's selected tumor and treatment goal could differ from the investigator's selection. If multiple tumors were associated with the primary palliative or preventive treatment goal and one tumor could not be singled out as the MTT, then the largest tumor was identified as the MTT.

Patients were treated (Appendix II) with 0.25 mL gel/cm³ of tumor volume weekly for 6 treatments within an 8-week period (Blinded Treatment Phase) or until patient objective complete response, whichever occurred first. Patients were evaluated weekly for 4 weeks after the last treatment and this completed the blinded treatment phase. Complete or partial responders were followed monthly for an additional 5 months or until disease progression. Patients who maintained a complete response at the end of 5 months continued to be observed monthly in the Extended Follow-up Phase (open-label). Patients with a partial response or relapse and those who required re-treatment in other tumors at the end of the treatment phase could enter the Extended Follow-up Phase and be retreated, if in the opinion of the investigator the patient could benefit from treatment. Also at treatment visit 4 or beyond, patients with progressive disease of the MTT could enter extended follow-up to receive open-label Intradose for a total of up to 6 treatments.

The Treatment Goal Questionnaire for palliation was required to be administered to the patient and the investigator at the screening visit and the beginning of every subsequent visit. The Treatment Goal Questionnaire for prevention was administered to the investigator at the screening visit, last evaluation (week 4) visit, last follow-up (month 5) visit, and last visit on study.

Reviewer's Comments:

1. Per original protocol, the assigned dose was 0.5 mL of CDDP/epi gel or placebo gel per cm³ tumor volume per treatment. However, the protocol was amended (Amendment V, May 2, 1997) for safety reasons to reduce the assigned dose to 0.25mL/cm³. At the time of amendment 72 patients were enrolled in strata 1 & 2 of Studies 414 and 514. Furthermore, the assigned dose for each treated tumor was recalculated at each treatment visit based on tumor volume at the time following the Amendment V, where as, prior to the amendment, tumor dose was calculated using the tumor volume at first treatment. Tumors smaller than 0.5 cm3 received a fixed dose of 0.1 mL per treatment.
2. This change of dosing schedule was not taken into account in the primary efficacy analyses of response data.

The study objectives of both the studies 414 and 514 were: 1) To compare the effect of CDDP/epi gel to placebo gel on local tumor volume. 2) To assess achievement of an identified primary treatment goal selected for the most troublesome tumor (identified by the investigator) in patients with recurrent or refractory squamous cell carcinoma of the head and neck following up to 6 weekly intratumoral treatments of CDDP/epi gel compared to placebo gel. 3) To compare the effect of CDDP/epi gel to placebo gel on total local tumor volume per patient. 4) To evaluate the time to response and the time to progression for MTT after local treatment with CDDP/epi gel as compared to placebo gel. 5) To assess improvement or stabilization in quality of life as measured by FACT-H&N. 6) To compare the histophathology of injected lesions that respond to local treatment. 7) Pharmacokinetics: To determine plasma platinum levels in patients receiving CDDP/epi gel.

A total sample size of 120 patients (80 evaluable patients in CDDP/epi gel treatment group and 40 patients in placebo gel treatment group), was computed in each of the Studies 414 and 514, which provides power of 0.8 or greater to detect a difference in response rates conditional on most troublesome tumor volume, using a two-sided alpha of 0.05, and based on the following assumptions: 1) Patients would be stratified by most troublesome tumor volume. The number of patients in each substrata would not be less than 20 and more than 30; 2) Response rate would be 10% or less in substrata treated with placebo gel; 3) Response rates in substrata treated with CDDP/epi gel would be between 40% and 70% in patients with MTT volume £ 5 cm³, and between 30% and 50% in patients with MTT volume > 5 to £ 20 cm³; 4) Response rates in patients with MTT volume > 20 cm3 would be 20% or less. The primary efficacy analyses of tumor response would be performed excluding this stratum. Furthermore, it was estimated that 88 patients would be needed to provide 80 evaluable patients in the CDDP/epi gel treated group, and 44 patients would be needed to provide 40 evaluable patients in the placebo gel treated group.

Reviewer's Comments:

1. The sample size calculations were based on testing the hypothesis of detecting a 10% difference in MTT objective response rate between the two treatment groups. No hypothesis was stated in the protocol with respect to patient benefit or the treatment goal (palliative or preventive) response.
2. Study MP-414-94-2 enrolled a total of 110 patients. Eighty-seven patients (as opposed a planned sample size of 132 patients) were enrolled in Strata 1 and 2 (CDDP/epi gel, 62 (31 in each stratum); placebo, 24 (12 in each stratum). Thus the actual patient allocation was 2.6:1 rather than the planned 2:1 per protocol. The 23 patients enrolled in Stratum 3 (CDDP/epi gel, 14; placebo, 9) were not part of the core analysis. The sponsor excluded one patient (ID # 2231 stratum 2, placebo) who did not receive study drug. This patient after starting the injection at the first treatment visit had to be withdrawn immediately due to severe pain. This patient was included in the safety analysis. Thus, a total of 86 patients in Stratum 1 and 2 were included in the efficacy analysis.
3. Study MP-514-94-2 enrolled a total of 116 patients. Ninety-two patients (as opposed a planned sample size of 132 patients) were enrolled in Strata 1 and 2 (CDDP/epi gel, 57 (31 in stratum 1 and 26 in stratum 2); placebo, 35 (17 in stratum 1 and 18 in stratum 2)). Thus the actual patient allocation was 1.5:1 rather than the planned 2:1 per protocol. The 24 patients enrolled in Stratum 3 (CDDP/epi gel, 17; placebo, 7) were not part of the core analysis. The analyses excluded one patient (ID# 4089 stratum 3, placebo) because this patient died the day after the screening visit and therefore was never treated. The total number of patients who were included in the efficacy analysis of this study was 92.
4. Subsequent to recommendations by the agency, a second primary efficacy variable (patient benefit) was added post-hoc. The understanding of the agency has been that the purpose of this trial is to demonstrate effects on both the designated primary variables, and hence adjustment of the type I error for multiplicity was not expected and therefore, both primary hypotheses could be tested at 0.05 level of significance (faxes sent by the Agency on (1) March 8, 2000 to sponsor in response to March 24, 2000 correspondence/request for feedback, and (2) April 28, 1998 reviewer comments on sponsor's submission of Sn 115, Feb. 24,1998).
5. No interim analysis was planned in both the studies.

3. Randomization

Each investigator received a block of patient numbers corresponding to a centralized randomization. Patients in stratum 1 were assigned numbers in a sequential order (from lowest to highest). Patients in stratum 2 were assigned numbers in reverse order (from highest to lowest). A separate block of randomized numbers were provided for all patients in stratum 3. Patients in stratum 3 were assigned numbers in sequential order. A patient number was assigned once that patient had completed all screening procedures and was deemed eligible for study entry.

Reviewer's Comments:

1. The following patients were reclassified into a different stratum post-randomization. The final analyses of the data were conducted based on the reclassified stratum allocation. All patients including reclassified patients in stratum 3 were excluded from efficacy analyses.

The primary efficacy endpoint, per original protocol, was objective response to treatment as measured by change in tumor volume for the MTT (volume 3.172, Section 4.3, page 18).

A complete response (CR) was defined as total disappearance of all clinically detectable and evaluable malignant disease, maintained for at least 4 weeks (25-28 days). A partial response (PR) was defined as 50% or greater reduction in detectable and evaluable malignant disease, maintained for at least 4 weeks (25-28 days). A stable disease was defined as < 50% reduction in detectable and evaluable malignant disease, or increase of < 25% detectable malignant disease and no significant change in measurable disease. A progressive disease was defined as 25% or greater increase in detectable and evaluable malignant disease. The overall objective response rate in each treatment arm was computed by combining the complete and partial responses. All the responses were evaluated with respect to the pre-selected MTT only.

Another endpoint that was designated as co-primary endpoint subsequent to agency's request (12/3/97 meeting minutes), was achievement of identified primary treatment goal and unforeseen additional benefit, as determined by the investigator and the patient. The investigator was required to select an improvable primary treatment goal for the patient's MTT from one of the following palliative or preventative categories: wound care, pain control, ability to see, hear or smell, physical appearance, obstructive symptoms, mobility, prevention of obstruction, invasion, or subcutaneous tumors breaking through the skin. The patient could also identify an improvable primary treatment goal (palliative only) for a MTT. Patient's selection of the MTT and treatment goal could differ from the investigator's selection. The achievement of treatment goal was measured by administering the Treatment Goal Questionnaire (Appendix I) to the investigator and the patient. The palliative goal was measured on a scale of 1-4 and the preventive goal was measured on a scale of 0-1 (met or unmet) (Appendix 1).

The following secondary efficacy endpoints were also evaluated in the two studies:

1. rate of objective response of all tumors treated during the blinded treatment phase,
2. duration of and time to response of all tumors treated during the blinded treatment phase,
3. time to progression of all tumors treated during the blinded treatment phase,
4. improvement or stabilization in quality of life as measured by the Karnofsky Performance Status,
5. improvement or stabilization in quality of life as measured by the FACT-H&N.

Reviewer's Comments:

1. The agency recommended strongly to the sponsor to consider symptomatic response as the primary efficacy endpoint in order to evaluate patient benefit from the proposed treatment (12/3/97 meeting with the sponsor and fax communication by the Agency to the sponsor on April 28, 1998 reviewer comments on sponsor's submission of Sn 115, Feb. 24,1998). Furthermore, the agency recommended to analyze the association of treatment goal attainment and MTT response. The Patient Benefit was to be based on the attainment of prospectively chosen primary treatment goals. Subsequently, the sponsor submitted an analysis plan in which patient benefit was described as an improvement of one scale point or more from baseline in the treatment goal. The agency expressed to the sponsor that a change in one scale point might not be sufficient to determine patient benefit (10/28/99 meeting).
2. The agency also recommended that the sponsor should analyze tumor response and patient benefit as co-primary endpoints (fax communication by the Agency to the sponsor on March 8, 2000 to sponsor in response to March 24, 2000 correspondence/request for feedback).
3. Achievement (post-hoc definition) of the prospectively identified primary treatment goal required either: i) an improvement of one scale point or more from baseline in the investigator or patient selected primary palliative treatment goal sustained for at least 4 weeks (25-28 days) or ii) absence of failure of the investigator- defined primary preventive treatment goals sustained for at least 4 weeks (25-28 days). Failure of the palliative primary treatment goal was defined as an increase in score of one point or more for two or more consecutive visits.
4. It is to be noted that the patient benefit was defined post-hoc as a change of one scale point. Furthermore, this change has not been validated prospectively to correspond to a significant patient benefit. Also, the palliative and prevention treatment goals are measured on different scales. The sponsor did not pre-specify the method of combining the multiple measurements in the protocol, and an interpretation of the resulting scale was not provided in terms of the size of a clinically relevant benefit.

5. Interim Analysis

Per protocol no interim analysis was planned.

Reviewer's Comments:

The minutes of the 12/3/97 meeting of the agency with the sponsor indicates that a blinded interim analysis was performed. The DSMB reviewed sequential unblinded analyses of data at intervals of approximately six months. To enable these analyses and remain blinded, Matrix used the services of an intermediary (Covance Clinical and Periapproval Services, Inc.).

Following an organizational meeting on 28-Jan-97, the DSMB held six meetings between March 1997 and November 1999 at which it reviewed unblinded data in closed session.

After each session, the DSMB announced its recommendations about whether study enrollment should continue. It also made recommendations for the safe participation of study patients. When the last patient completed the Treatment Phase and the study itself was unblinded, Matrix obtained the minutes of the six data review meetings. A final

DSMB meeting was held on 21-Aug-00, when the unblinded results were openly discussed with Matrix for the first time.

No interim analyses were conducted by Matrix. Matrix remained blinded to patients’ treatment assignments until all patients had completed the Treatment Phase and all results for objective response and outcome of treatment goals were entered and locked in the database. No changes in objective response or outcome of treatment goals were made after the study was unblinded.

Patients who received at least one treatment were to be considered evaluable for the intent-to-treat efficacy analysis. Patients who received six treatments to the MTT, patients who had a complete response for the MTT in less than six treatments, and patients who withdrew from treatment early for a study-related reason were to be evaluable for a per-protocol (efficacy analyzable fraction) analysis. Only tumors measuring 0.5 cm³ or larger were considered evaluable for efficacy. At the time the protocol was amended to include patients with tumors greater than 20 cm³ (Stratum 3), it was noted that these patients would not be part of the efficacy analysis.

Analysis of MTT response was to be performed for patients in Strata 1 and 2 only.

Patients in Stratum 3 were not to be included in the primary efficacy analysis. Response of the investigator-selected MTT was to be determined by the maximum decrease in MTT volume between baseline and four weeks after the end of treatment (or the last measured volume for patients who terminated early). The mean, median, standard deviation, and range for absolute and percent change in most troublesome tumor volume were to be presented for all patients by stratum (excluding Stratum 3) and treatment group, and, as sample size permitted, by age, sex, disease history, and treatment history. The best percent change in MTT volume at four weeks post-treatment, conditional on baseline tumor volume, was to be calculated and compared across treatment groups using ANOVA or ANCOVA. A secondary analysis was to assess the difference between treatment groups at the time of best response, i.e., when the maximum decrease of MTT volume occurred. Each patient’s best MTT response at the end of Follow-up (or at early termination) was to be recorded.

The percentage of patients with a CR, PR, SD or PD as best response of the MTT were to be compared across treatment groups conditional on baseline tumor volume stratum, using the Mantel-Haenszel chi-squared test. In addition, comparisons of CR + PR (responders) vs. SD + PD (non-responders) were to be made across treatment arms, adjusted for tumor volume baseline stratum, using the Mantel-Haenszel chi-squared test. Population response rates (for CR and CR + PR) for both treatment groups were to be estimated using two-sided 95% confidence intervals. Where sample size permitted, inter-group comparisons of MTT response were also to be made for subgroups defined by age, sex, disease history, and treatment history.

Time to CR of the MTT was to be compared across treatment groups using the Kaplan-Meier method with baseline total tumor volume as a covariate. Time to CR + PR were to be similarly compared. In all time-to-response analyses, patients who failed to respond, including early terminators, were to be censored. Recurrence of disease after CR was defined as the return of the MTT or the incidence of a new tumor or tumors in the treated area ("treated area" as defined by the investigator). Duration of response was defined as the time from the beginning of a CR of the MTT to recurrence. Time to recurrence (equivalent to duration of response) was to be calculated for each CR as precisely as the visit schedule allowed. Time to recurrence following CR was to be compared between treatment groups using the Kaplan-Meier method, which would also provide estimates of recurrence rate by treatment group.

Disease progression of the MTT was defined as a 25% or greater increase in MTT volume from any previous measurement. Progression could occur at any time after the first treatment, and could be preceded by a CR or PR. Time to disease progression was to be compared across treatment groups using the Kaplan-Meier method. The Follow-up and Extended Follow-up periods were to be designed to enable estimation of time to disease progression for all patients except those having very durable responses. The patient responses of patients initially treated with placebo and subsequently treated with open-label CDDP/epi gel in Extended Follow-up were to be contrasted using a paired design.

The statistical methods used for the MTT response were also to be used to evaluate total patient response, i.e., the change in volume of all treated tumors. Responses in treated tumors other than the MTT were to be assessed descriptively. The number and percent of tumors classified as CR, PR, SD or PD at the end of the Treatment Phase were to be tabulated by treatment group.

Achievement of the palliative primary treatment goal selected by the patient and investigator for the MTT was to be assessed by examining trends in scores on the Treatment Goal Questionnaire (TGQ) over the Treatment, Follow-up, and Extended Follow-up Phases. At each patient visit (Screening, Treatment, Follow-up) the scores for the selected treatment goals were to be recorded. Scores from screening and Visit 1 were to be averaged to obtain a baseline primary treatment goal score. Within each treatment goal category, scores for the primary treatment goal were to be compared across treatment groups at the end of the treatment period using analysis of variance (ANOVA) or analysis of covariance (ANCOVA) with baseline total tumor volume as a covariate. Other treatment goals selected by the investigator and patient were also be scored at each visit and compared across treatment groups within treatment-goal categories. Mean per-patient scores for all identified goals and scores for all primary treatment goals across all categories were also to be examined.

Medians and ranges for goals scores by category and overall were to be represented graphically by treatment group over time, to identify trends during Treatment, Follow-up, and Extended Follow-up Phases. If appropriate, the General Estimating Equations method was to be used to examine the long-term effect of treatment on the primary treatment goal and other individual treatment goals. The effect of baseline tumor volume and treatment history on achievement of palliative treatment goals was to be examined. Other analyses of treatment goal outcomes were to include the examination of treatment goal scores at the time of maximum response for the MTT and repetition of the analysis described above for the subgroup of patients classified as having complete response of the MTT at the end of the treatment period, and for those having complete patient response at the end of study.

The reliability of the Palliative Treatment Goal questionnaire was to be investigated by comparing the Screening and Visit 1 responses using McNemar’s test. An attempt was to be made to validate the questionnaire by examining the association of change in

questionnaire score for the primary treatment goal with changes in volume of the MTT and changes in FACT-H&N score occurring during the same time period. The physician options for treatment goals also included prevention. Performance on preventive goals was to be assessed at the fourth weekly visit following the end of randomized treatment, the 5-month Follow-up visit, and study completion. The proportion of patients failing the preventive goal at each evaluation timepoint was to be compared across treatment groups for all prevention categories combined and, if sample sizes permitted, within specific categories, using the Mantel-Haenszel chi-squared test. If possible, the effect of baseline total tumor volume and treatment history on the success of preventive treatment goals was to be examined. An attempt was to be made to validate the preventative goal section of the questionnaire by examining the association of prevention/non-prevention with changes in tumor volume for the most troublesome tumor and changes in the FACT-H&N score occurring during the same time period. The extent of agreement between patient and physician was to be evaluated in the sub-sample of patients, in which the patient and investigator selected the same goal, using the kappa statistic.

To evaluate the impact of treatment on global quality of life, each patient was to complete the FACT-H&N quality of life assessment regularly while on study. This instrument included a subscale specific to head and neck cancer. FACT-H&N responses at each visit were to be scored as described in the FACT Manual. Missing responses were to be pro-rated as permitted by the Manual. Individual scores obtained on the six subscales, and the overall FACT-H&N score, were to be summarized and examined by visit, by the response status of the MTT, and by treatment group. Nonparametric analysis of variance and tests for trend were to be used to identify significant differences between treatment groups in FACT-H&N overall and subscale scores over time. Changes from baseline in FACT-H&N scores were to be compared across treatment groups at the 1-week and 4-week post-treatment visits, the 5-month Follow-up visit, and study completion. The associations between subscale and overall FACT-H&N responses and KPS score over the study period were to be examined using Spearman's correlation coefficient.

• The response of other treated tumors relative to the response of the MTT to be examined;

• A logistic regression model, exploring the relationship between MTT response and baseline patient and disease characteristics to be provided;

• Analysis of the relationship between MTT response and Patient Benefit was expanded to include exploration of the reasons patients maybe classified as benefitters in the absence of MTT response.

Pairs of binomial proportions (for instance, Stratum 1 CDDP/epi gel and placebo MTT response rates) were compared using Pearson’s (large-sample) chi-squared test and

Fisher’s exact test, as implemented in SAS procedure FREQ. Fisher’s exact test was also used to compare CDDP/epi gel rates across patient subgroups: for instance, MTT response rates across centers. For stratified comparisons, the hypothesis that CDDP/epi gel and placebo rates are equal within each stratum was tested using large-sample and exact Cochran-Mantel-Haenszel tests. The large-sample approximations were obtained using FREQ and the exact p-values were obtained using SAS procedure MULTTEST (code supplied by Peter Westfall, one of the authors). Exact Clopper-Pearson confidence intervals for binomial rates were computed. Cramér’s V statistic, which for 2x2 contingency tables is the signed square root of [Pearson statistic / n], was used as a measure of association of binary variables, such as MTT response (responder, non-responder) and Patient Benefit (benefitter, non-benefitter).

SAS procedure GENMOD was used to fit an exploratory no-interaction logistic regression model to the blinded-phase MTT response status of CDDP/epi gel patients in Stratum 1 and Stratum 2. Variables such as age and time from original diagnosis of primary disease to first treatment visit were categorized in the same way they were categorized for the corresponding one-variable-at-a-time MTT response summary tables.

With the intention of conducting analysis based on combined-studies results, cutpoints for continuous variables were chosen to be approximately equal to tertiles or quartiles of the combined distribution of CDDP/epi gel and placebo groups. The no-interaction model and one-variable-at-a-time models were fitted solely to explore which variables should be included in the combined-trials model and the results should not be used as a basis for conclusions about which factors predict response.

Three types of time-to-event ("survival") summaries were computed. Where possible, time-to-event distributions were estimated by the Kaplan-Meier method, as implemented in SAS procedure LIFETEST. Statistics such as estimated median time to event, estimated mean time, and the estimated standard error of the mean estimator are those calculated by LIFETEST with the default options. Time-to-event summaries were also computed for exact times only. For instance, time to MTT response was summarized for MTTs with CR or PR, and time to MTT response for CDDP/epi gel responders was compared across patient subgroups using Wilcoxon’s rank sum tests or Kruskal-Wallis tests, as implemented in SAS procedure NPAR1WAY, using the EXACT statement where feasible.

1. Although the primary analysis was conducted in all the randomized patients to strata 1 and 2, the data was analyzed per re-classification of some of the patients to different strata (see reviewer's comments in section 2.2.3 above).
2. It appears from the statement underlined in the analysis plan specified above, that the Treatment Goal Questionnaire is not prospectively validated instrument for measuring a meaningful clinical patient benefit. The sponsor's submission of the validation of the treatment goals questionnaire (vol. 3.14) is based on 12 investigators and 15 patients from one single institution, who were not entered in the trial. None of these 15 patients selected pain as the most important treatment goal (vol. 3.14, page 28). However, patients in the current studies under review most commonly (34%) selected pain as the primary or most important treatment goal.
3. Furthermore, it appears that in the initial plan of analysis, the palliative and preventative goals were to be assessed and validated separately with respect to FACT-H&N scores. However, in the NDA application an algorithm was presented where in the results of the two goals are combined as presented below.
4. Plan for comparing treatments with respect to patient benefit when there is differential number of treatments in the CDDP/epi gel and Placebo gel was not discussed.
5. Analysis plan accounting for the change in dose assignment after Amendment V is not specified.

Sponsor's Algorithm for Evaluating Patient Benefit

MTT volume, number of tumors at the start of the study, treatment history, baseline Karnofsky performance status, FACT H&N score, and age were recorded at the study entry. There appears to be no significant imbalance with respect to these baseline factors between the CDDP/epi gel arm and the placebo arm. Demographic and other baseline characteristics are summarized in Table 1a and 2a, respectively, for Studies 414 and 514.

Reviewer's Comments:

1. The mean and median MTT volume in stratum 1 is larger in the Study 514 compared to Study 414.
2. Median MTT volume in the placebo arm was smaller than the CDDP/epi arm in Stratum 2 of Study 514.
3. The summary of demographic data presented in Tables 1b and 2b, respectively, for studies 414 and 514, are as per the original randomization scheme (for difference in patient numbers in the two strata compared to Tables 1a and 1b, see reviewer's comments, section 2.2.3). There appears to be no apparent difference between the two sets of data (Table 1a vs. 1b and Table 2a vs. 2b).

1. Primary Efficacy Evaluation

The following Tables 3a & 4a, compare the objective tumor (CR + PR) response rate per Sponsor’s evaluation during the blinded treatment phase of the study, respectively, for Studies 414 and 514. During the blinded period, rates of MTT response of the CDDP/epi gel arm and the placebo arms per sponsor's analyses were significantly different. In Study 414, 34% of patients with CDDP/epi gel had tumor responses while no patient with placebo had response (p= 0.001). In Study 514 the tumor response rates in the CDDP/epi gel and placebo gel arms were, respectively, 25% and 3% (p= 0.007).

* Cochran-Mantel-Haenszel test (exact)