Frequently Asked Questions
Data set
Q: What are individual participant data (IPD)?
A: IPD are the timing of events for individual patients enrolled in a clinical trial. Event types most often include censoring or death for cancer clinical trials, but can also include surrogate events such as disease progression, biochemical or clinical treatment failure, and local-regional recurrence of disease.
Q: How does the data imputation procedure work, and how do you check the quality of the resulting data?
A: The data extraction and imputation procedures followed the methods outline in the following two studies:
- Guyot, P., Ades, A., Ouwens, M.J., and Welton, N.J. (2012). Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves. BMC Med Res Methodol 12, 9.
- Rahman, R., Fell, G., Ventz, S., Arfé, A., Vanderbeek, A.M., Trippa, L., and Alexander, B.M. (2019). Deviation from the Proportional Hazards Assumption in Randomized Phase 3 clinical Trials in Oncology: Prevalence, Associated Factors, and Implications. Clin Cancer Res 25, 6339–6345.
The quality of the data imputation was confirmed quantitatively by calculating the Hazard Ratio using this imputed data and compared to the corresponding trial’s reported Hazard Ratio and qualitatively by overlaying the KM curve generated from the imputed data on top of the published curve. Trials with a Hazard Ratio difference greater than 0.1 or perceptible visual differences, were removed from the final data set and not analyzed further.
Q: How many studies are included in the current data set?
A: The data set consists of ~150 unique published phase III clinical trials in breast, colorectal, lung, and prostate cancer in metastatic and non-metastatic setting from 2014-2016. The resulting curated data set consists of 262 IPD .csv files, each with data from a unique image from the published trial results. In aggregate, these data comprise ~ 220,000 overall survival or event-free survival events (e.g. progression free survival, PFS).
Analysis
Q: Is this study a systematic review or meta-analysis?
A: No. A systematic review is a formal study design to gather data, based on prespecified eligibility criteria, to answer a specific question. A meta-analysis is a subtype of systematic review that integrates the results of previous research studies to make conclusions relevant to that disease and treatment area, such as to provide a pooled estimate of a treatment’s effect.
Our work performed exploratory analysis on cancer survival data broadly across many types of oncology clinical trials and treatment modalities. We do not intend to make specific conclusions about a single treatment or therapy. None of our analysis is intended to influence treatment decisions.
Q: How is the parametric fitting procedure performed?
A: The event times for the imputed patients, either death for overall survival distributions or surrogate events in the case of event-free survival distributions, are compared to the event times simulated under each parametric distribution. The likelihood of a specific parametric form to fit patient data is computed by maximum likelihood estimation. Specifically, the relative likelihood of a patient event taking place at a particular point in time is calculated under that parametric distribution’s probability density function. The likelihood of a censoring event taking place is calculated by integrating the probability density function (the cumulative density function), and computing the likelihood of a patient event taking place in the trial after the censoring time (1- the probability at that time under the cumulative density function). This procedure is repeated for all patient events in an arm of a clinical trial, and the overall likelihood of a fit was calculated by multiplying all relative likelihoods.
Q: Why does the parametric fit look worse at the end of a trial than at the start?
A: Because this is the region of the survival curve with the fewest number of patient events. The fitting procedure maximizes the likelihood of a single distribution to describe all observed patient events, and therefore its best performance will be at the regions of the survival curve with the most data.