Time Dependent Surrogates as Predictors of Overall Survival in Multiple Mieloma

Background: Time to progression (TTP), progression-free survival (PFS) and event-free survival (EFS) are common surrogates in clinical cancer investigation and acceptable end points informing decisions about new drugs approval and financing. Our aim was to estimate a quantitative relationship between median TTP, PFS and EFS and median overall survival (mOS) in multiple mieloma (MM) from data of prospective (experimental or observational) studies published in the literature.

Methods: Studies published in English, Spanish or Portuguese between 1970 and 2007 were systematically searched on PubMed using keywords: progression, event-free, survival, multiple myeloma, clinical trial and observational study. All types of treatments were considered with the exception of allogeneic transplantation. The non-parametric Spearman’s rank correlation coefficient was used as a measure of correlation between median values of the surrogates and median values of OS. The quantitative relationship between surrogates and OS was estimated with a two-step approach to a simultaneous Tobit model. Study arms not reaching median OS were included as censored observations. First the endogenous variable (TTP/PFS/EFS) was regressed on the instrumental variable (overall survival at 12 months) and on the exogenous variables (median age, percentage of females, year of publication, type of surrogate, and the type of patients included in the trials naïve vs non naïve) using Generalized Method of Moments - Cragg estimator to obtain a consistent estimate of the residuals. Second the censored normal model was estimated by maximum likelihood including the estimated residuals as an additional regressor. Estimation was weighted by the number of patients enrolled in each study arm and estimators’ variances were corrected for both endogeneity and heteroskedasticity.

Results: Of the 845 studies reviewed 128 were included containing a total of 190 arms: 34 reported TTP, 71 reported EFS, 85 reported PFS, and 142 reported mOS. The mean duration of these studies was 4.1 years. Overall, the sample was composed of 17163 patients, 56% males, mean age (mean of medians) was 60 years, with 52% of arms representing only naïve patients. Mean of median TTP/PFS/EFS was 23 (SD=16) months and mOS was 39 (SD=19). The correlation coefficients of median TTP, PFS and EFS with mOS were 0.48 (p=0.01), 0.74 (p<0.0001), and 0.84 (p<0.0001), respectively. The model estimates a 1.9 (95%CI [1.5;2.2]) months increase in median overall survival for each additional month reported for the median time-dependent surrogate. There was no evidence that this relationship differs by the type of surrogate, when controlling for age, gender, type of patients included in the trials (naïve vs non naïve) and year of publication.

Conclusions: The analysis confirms the value of time dependent surrogates (TTP/PFS/EFS) in predicting overall survival in patients with Multiple Myeloma. The quantitative relationship presented is of most value to inform clinical trials’ design and to support new multiple mieloma drugs approval and financing decisions.