Association between pertuzumab induced diarrhea and cutaneous rash and survival outcomes in patients with HER2-positive metastatic breast cancer enrolled in the CLEOPATRA trial

a.Cohort definitionFrom the overall CLEOPATRA study cohort we will define a sub-cohort of patients evaluable for this analysis, the “rash-diarrhea evaluable cohort”. This sub-cohort is composed of all patients free of progression/death at date of docetaxel withdrawal plus 46 days and aims at 1) focus on the specific effects of the biologic component of treatment (pertuzumab + trastuzumab), and 2) addressing the risk for immortal person-time bias. Given that the study objectives aim at associate pairs of outcomes, e.g. diarrhea and PFS, there is the risk of immortal person-time bias, because the defining feature of the cohort (in this example having developed diarrhea or not) is not a baseline characteristic, but an outcome in itself. Therefore, patients that will develop diarrhea are “protected” from, in this example, progression or death while not developing diarrhea because, until then, they will integrate the group of patients not developing diarrhea. This generates the commonly referred immortal person-time bias because the group not developing diarrhea will be enriched of patients with early disease progression or death. To deal with this risk of bias, a guarantee time of 46 days free of progression/death after docetaxel withdrawal will be a selection criterion for the definition of the “rash-diarrhea evaluable cohort” and thus as the defining rule for a landmark analysis. Landmark analysis has previously shown to be useful in dealing with this risk10 and a related study successfully used this approach.11 The time to docetaxel withdrawal was selected because we are focused on the adverse events specifically associated with the biologic combination, not chemotherapy. In addition, the 46 extra days account for the longest median time to the development of rash/diarrhea in several studies using pertuzumab (longer interval to rash/diarrhea occurred in the placebo arm: median to rash of 46 days [P25-P75 2-847 days] and median to diarrhea of 23 days [P25-P75 6-68 days]).6,7b.Cohort characteristicsDemographic, clinico-pathological and treatment characteristics will be tabulated for the overall cohort and according to: 1) development of rash, and 2) development of diarrhea. c.Association between rash, diarrhea and ageA two-by-two table will be created tabulating the development of rash and diarrhea, by age (50 years cut-off), overall and according to treatment arm. Chi-squared test will be performed. The 50 years cut-off was chosen given that: 1) it is close to the median age of the study population (median in both arms of 54.0 years), which assures a split of the overall population in two reasonably well balanced samples, and 2) it is commonly used as an indicator variable/surrogate of the menopausal status as it slightly above the median age of menopause in several western countries.d.Association of diarrhea/rash with survival outcomesFor time to event outcomes, we will plot survival rates using Kaplan-Meyer (KM) plots and compare differences in survival rates using univariate and multivariate Cox proportional hazards models (Cox models). In the group of patients receiving pertuzumab, we will perform KM plots of survival outcomes according to the occurrence of diarrhea (overall and by grade [<3 and =3]) and of skin rash (of any grade) for the overall cohort and according to age at diagnosis of metastatic disease (=50 years and >50 years). A multivariate Cox proportional hazard model will be fitted including (tentatively):• Independent variables: diarrhea (as “yes/no” and as “yes grade <3, yes grade =3, no”), and rash;• Covariates: age, hormone receptor status, EGFR mRNA levels, ECOG performance status, type of metastatic disease (de novo vs. recurrent), type of metastatic site of disease (visceral vs. non-visceral), previous (neo)adjuvant chemotherapy, and previous (neo)adjuvant anti-HER2 therapy; • Outcomes: PFS (by independent and investigator assessment) and overall survival;• Interaction terms: age and diarrhea, age and rash, age and diarrhea severity, EGFR mRNA levels and diarrhea, EGFR mRNA levels and rash, EGFR mRNA levels and diarrhea severity.e.Efficacy of pertuzumab according to diarrhea/rashIn the overall study cohort (i.e., including both treatment arms), we will perform KM plots of survival outcomes according to treatment arm and by the occurrence (vs. not) of diarrhea and of skin rash for the overall cohort and according to age at diagnosis of metastatic disease (=50 years and >50 years). Multivariate Cox proportional hazard models will be fitted for each case including (tentatively):• Independent variables: treatment arm, diarrhea, and rash;• Covariates: age, hormone receptor status, EGFR mRNA levels, ECOG performance status, type of metastatic disease (de novo vs. recurrent), type of metastatic site of disease (visceral vs. non-visceral), previous (neo)adjuvant chemotherapy, and previous (neo)adjuvant anti-HER2 therapy; • Outcomes: progression-free survival (by independent and investigator assessment) and overall survival;• Interaction terms: treatment arm and diarrhea, and treatment arm and rash.f. Predictive model for survival outcomesWe will build a predictive model for survival outcomes (reduced model) and subsequently extend this model to include the occurrence of diarrhea and rash (extended model). Likelihood ratio tests will be used to test the hypothesis that there is a significant difference between a model including or not diarrhea/rash. Tentative reduced model includes:• Independent variables: treatment arm;• Covariates: age, EGFR mRNA levels, hormone receptor status, ECOG performance status, type of metastatic disease (de novo vs. recurrent), type of metastatic site of disease (visceral vs. non-visceral), previous (neo)adjuvant chemotherapy, and previous (neo)adjuvant anti-HER2 therapy; • Outcomes: progression-free survival (by independent and investigator assessment), and overall survival;Extended model further includes:• Independent variables: occurrence of diarrhea, and occurrence of rash;• Tentative interaction terms: age and diarrhea, age and rash, EGRF mRNA levels and diarrhea, EGFR levels and rash.Proportional hazards assumption will be verified by testing for a nonzero slope in a linear regression of the Schoenfeld residuals versus time. Given the randomized nature of the cohort, missing data will be considered as missing at random. Given that no substantial lost to follow-up or missing data are expected for demographic, clinicopathological characteristics and adverse events (see details in the “Power and sample size” section), for model construction, patients with missing data in variables included will be excluded from the analysis. All tests will be two-tailed and p-value <0.05 will be considered significant. This statistical analysis plan is subject to revision when we will have the opportunity to examine data availability.Power and sample sizeCLEOPATRA trial enrolled 808 patients, of which 402 in the pertuzumab group and 406 in the placebo group. In the last study report, 604 progression-free survival events were registered (284 in the pertuzumab arm and 320 in the control arm). From the available data, we estimate that up to 20% of patients will not be included in the “rash-diarrhea evaluable cohort” (docetaxel withdrawal plus 46 days of treatment). No substantial lost to follow-up or missing data are expected for demographic, clinicopathological characteristics and adverse events, as reported in the two main publication detailing CLEOPATRA results.3,12 Specifically, from the overall group of characteristics and outcomes assessed, only 13 adverse events have an undetermined status/are missing, of which in one patient for rash status (if overall missing cases refer to individual patients or are clustered in less individuals is not detailed in the manuscript). Overall, after docetaxel suspension, diarrhea occurred in 21.7% of patients (28.1% in the pertuzumab group and 14.2% in the control group), while rash occurred in 13.6% (18.3% in the pertuzumab group and 8.0% in the control group) of patients. For a significance level of 5% and an expected impact of the development of diarrhea or rash on PFS of 40% (hazard ratio of 0.6013), we estimate a power of 99.2% and 94.6% to detect an association of the development or diarrhea and rash with diarrhea, respectively.