Finding an Effective Dose of GM1 to Reduce or Prevent Neuropathy (Numbness or Weakness) Due to Treatment With Paclitaxel (Phase II)

Defined as the highest dose level that induces dose-limiting toxicity in less than one-third of patients (less than 1 out of 3 or less than 2 out of a maximum of 6 new patients).The constellation of adverse events as scored using the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events, version (v) 5.0 (CTCAE v5.0) will be summarized by reporting the number and percentage of patients. Specifically, the number and severity of all adverse events (overall and by dose-level) will be tabulated and summarized. The grade 3+ adverse events will also be described and summarized in a similar fashion. Further, all adverse events will be listed by system organ class and preferred term.

Reflects sensory paclitaxel-induced peripheral neuropathy symptom severity and onset. Measured using 6 individual Quality of Life Questionnaire-Chemotherapy-Induced Peripheral Neuropathy 20 (QLQ-CIPN20) questions that quantify numbness (N), tingling (T) and pain in the fingers/hands and toes/feet. Will calculate the highest (worst) N, T, and pain sensory score obtained anytime during paclitaxel exposure. Response is defined as a patient reporting a highest score of =< 2 without discontinuing the study due to sensory paclitaxel-induced peripheral neuropathy. Chi-squared test will be used. Will report the proportion of patients achieving a composite response in each arm as well as the difference in proportions along with the 90% confidence interval. Will report the estimated odds ratio (monosialotetrahexosylganglioside [GM1]/placebo) and the corresponding 90% confidence interval for the true, but unknown odds ratio. Multiple logistic regression will be used to estimate the intervention

Binary endpoint (yes or no). will graphically look at the percentages of patients receiving full planned paclitaxel doses per cycle within each arm. This is an indirect measure of paclitaxel-induced peripheral neuropathy, as most patients who stop receiving full dose paclitaxel do so because of paclitaxel-induced peripheral neuropathy troubles. In addition to graphically displaying the percentage of patients who receive full dose paclitaxel at each cycle, and according to arm, will make a between-arm comparison in terms of changes in the probability of receiving full dose paclitaxel over the 12 cycles of treatment. For latter analysis, the generalized estimating equations approach will be applied, with inference based on the empirical or sandwich variance estimator.

Serially measured. Item scores are totaled and linearly converted to a 0-100 point scale with higher scores representing fewer symptoms or better quality of life. Will use the area under the curve (AUC) to summarize the serially measured sensory subscale scores. Patients will be analyzed in the arms to which they were randomized. In order to adjust for the stratification factors, the AUC will be computed for the two arms based on estimated parameters from a repeated-measures (means) mixed model.

The proportion of patients experiencing a grade 3+ adverse events or toxicities will be described for each treatment arm, and will also be compared between the arms using Fisher's exact tests. Additional summaries and test will be performed as needed.

Within each arm, the mean score at each time point will be plotted longitudinally for the endpoints measuring aches and pains at its worst. The functional form of the relationship between time and each of these three patient-reported acute neuropathy items is not known. Penalized splines, a smoothing technique which does not require strong assumptions concerning the functional form of the pattern of change in the mean response, will be applied to these longitudinal data. In the two-arm setting (GM1 vs placebo) time trends will be incorporated in a non-parametric fashion, thereby allowing the mean response to change in a highly non-linear, but not predetermined, way. The intervention effect will be incorporated in the model in a parametric fashion, thereby allowing a relatively simple, but powerful, test of the intervention effect on changes in the mean response over time.

Within each arm, the mean score at each time point will be plotted longitudinally for the endpoints measuring aches and pains at its least. The functional form of the relationship between time and each of these three patient-reported acute neuropathy items is not known. Penalized splines, a smoothing technique which does not require strong assumptions concerning the functional form of the pattern of change in the mean response, will be applied to these longitudinal data. In the two-arm setting (GM1 vs placebo) time trends will be incorporated in a non-parametric fashion, thereby allowing the mean response to change in a highly non-linear, but not predetermined, way. The intervention effect will be incorporated in the model in a parametric fashion, thereby allowing a relatively simple, but powerful, test of the intervention effect on changes in the mean response over time.

Within each arm, the mean score at each time point will be plotted longitudinally for the endpoints measuring aches and pains on average. The functional form of the relationship between time and each of these three patient-reported acute neuropathy items is not known. Penalized splines, a smoothing technique which does not require strong assumptions concerning the functional form of the pattern of change in the mean response, will be applied to these longitudinal data. In the two-arm setting (GM1 vs placebo) time trends will be incorporated in a non-parametric fashion, thereby allowing the mean response to change in a highly non-linear, but not predetermined, way. The intervention effect will be incorporated in the model in a parametric fashion, thereby allowing a relatively simple, but powerful, test of the intervention effect on changes in the mean response over time.

Distributions of PFS times will be estimated using Kaplan-Meier methodology.

Distributions of OS times will be estimated using Kaplan-Meier methodology.

Will use the AUC to summarize the serially measured EORTC QLQ-CIPN20 scores. Patients will be analyzed in the arms to which they were randomized. In order to adjust for the stratification factors, the AUC will be computed for the two arms based on estimated parameters from a repeated-measures (means) mixed model.

As measured by the Patient Reported Outcomes version CTCAE (PRO-CTCAE). The maximum grade for each of these events will be recorded for each patient and frequency tables will be generated and presented according to treatment arm. The descriptive analysis of the PRO-CTCAE endpoints will be based on the safety population and analyzed according to the treatment actually received.

Measured by the 11-item Functional Assessment of Cancer Therapy - Gynecologic Oncology Group-Neurotoxicity 4. Will tabulate all available serially measured PROs at each protocol defined time point and within each arm; furthermore, will graphically display the longitudinal PROs over time. Will compare and contrast the descriptive findings between these CIPN-measuring instruments in this trial and in the corresponding longitudinal data obtained in S1714. No inferential statistics will be generated. Further, these descriptive summaries will be used to better understand these instruments.

Measured by the 5-item Patient Reported Outcomes Measurement Information System-Neuropathic Pain Quality scale. Will tabulate all available serially measured PROs at each protocol defined time point and within each arm; furthermore, will graphically display the longitudinal PROs over time. Will compare and contrast the descriptive findings between these CIPN-measuring instruments in this trial and in the corresponding longitudinal data obtained in S1714. No inferential statistics will be generated. Further, these descriptive summaries will be used to better understand these instruments.