Do Cancer Clinical Trial Populations Truly Represent Cancer Patients? A Comparison of Open Clinical Trials to the Cancer Genome Atlas

Nophar Geifman, Atul J. Butte

Introduction

In the era of precision medicine, researchers increasingly rely on open-access clinical trial data to validate hypotheses and conduct meta-analyses. However, a fundamental question remains: are the participants in these trials truly representative of the average patient seen in an oncology clinic? Clinical trials often employ strict inclusion and exclusion criteria, potentially "cherry-picking" subsets of patients—such as those with end-stage disease who have failed prior treatments—that do not reflect the broader patient population. This study quantifies this gap by comparing patient-level data from open clinical trials with The Cancer Genome Atlas (TCGA), which serves as a proxy for the general cancer population.

Methodology

The researchers performed an integrative analysis of patient-level data across four major cancer types: breast, colorectal, prostate, and pancreatic. The study utilized two primary datasets:

  • Project Data Sphere (PDS): A portal providing patient-level data from the comparator arms (standard-of-care groups) of Phase III cancer trials. Data from 10 trials (approx. 3,300 subjects) were used.
  • The Cancer Genome Atlas (TCGA): A repository characterizing the genetics and clinical outcomes of a wide range of cancer stages, serving as the "general population" baseline (approx. 2,000 subjects).

To ensure comparability, the team mapped disparate data schemas into a uniform representation. This involved manual staging for prostate cancer trials based on tumor (T) and metastasis (M) status. Survival analysis was conducted using Kaplan-Meier estimators and log-rank tests to determine statistical significance between groups.

Data processing and harmonization workflow for integrating clinical trial and TCGA datasets.

Results and Discussion

The Survival Gap and "Survival Flip"

The analysis revealed that survival rates in clinical trials were significantly lower ($p < 0.0001$) than in the TCGA cohort. This discrepancy is primarily attributed to the fact that trial populations are heavily skewed toward Stage IV (advanced) disease, whereas TCGA includes a broader mix of Stages I through IV.

A striking finding was the "Survival Flip." In the general population (TCGA), breast cancer patients typically exhibit better survival outcomes than colorectal cancer patients. However, in the clinical trial dataset, this relationship was reversed: colorectal patients showed better survival than breast cancer patients. This highlights how trial-specific selection criteria can fundamentally alter observed disease outcomes.

Survival analysis comparing clinical trial data (PDS) and the general population (TCGA) across four cancer types.

Demographic and Selection Biases

Clinical trials also demonstrated a clear age bias, underrepresenting elderly patients (over age 75) compared to the TCGA baseline. This suggests a selection bias toward younger, potentially "fitter" patients. Furthermore, even when comparing only Stage IV patients, trial subjects (except for colorectal) had lower survival than TCGA subjects. This "last resort" effect likely stems from trial participants having failed multiple lines of therapy before enrollment, whereas TCGA patients may be at the beginning of their treatment journey.

Colorectal cancer was the sole exception where Stage IV trial subjects matched TCGA survival rates. This was attributed to the specific trial analyzed being a first-line treatment study, making its population more comparable to the general clinical population.

Conclusion and Future Directions

This study demonstrates that clinical trial populations are often not representative of the broader cancer patient population, particularly regarding disease stage and prior treatment history. While open-access data is a powerful tool, its non-uniformity and inherent selection biases require careful consideration during analysis. Future work will involve expanding this analysis to over 48 trials available on PDS and integrating more complex variables, such as specific drug regimens and molecular subtypes, to further bridge the gap between trial data and real-world clinical outcomes.

Summary updated

Abstract

Open clinical trial data offer many opportunities for the scientific community to independently verify published results, evaluate new hypotheses and conduct meta-analyses. These data provide a springboard for scientific advances in precision medicine but the question arises as to how representative clinical trials data are of cancer patients overall. Here we present the integrative analysis of data from several cancer clinical trials and compare these to patient-level data from The Cancer Genome Atlas (TCGA). Comparison of cancer type-specific survival rates reveals that these are overall lower in trial subjects. This effect, at least to some extent, can be explained by the more advanced stages of cancer of trial subjects. This analysis also reveals that for stage IV cancer, colorectal cancer patients have a better chance of survival than breast cancer patients. On the other hand, for all other stages, breast cancer patients have better survival than colorectal cancer patients. Comparison of survival in different stages of disease between the two datasets reveals that subjects with stage IV cancer from the trials dataset have a lower chance of survival than matching stage IV subjects from TCGA. One likely explanation for this observation is that stage IV trial subjects have lower survival rates since their cancer is less likely to respond to treatment. To conclude, we present here a newly available clinical trials dataset which allowed for the integration of patient-level data from many cancer clinical trials. Our comprehensive analysis reveals that cancer-related clinical trials are not representative of general cancer patient populations, mostly due to their focus on the more advanced stages of the disease. These and other limitations of clinical trials data should, perhaps, be taken into consideration in medical research and in the field of precision medicine.
BibTeX
@inproceedings{geifman2016do,
  author = {Nophar Geifman and Atul J. Butte},
  title = {{Do Cancer Clinical Trial Populations Truly Represent Cancer Patients? A Comparison of Open Clinical Trials to the Cancer Genome Atlas}},
  booktitle = {PSB},
  year = {2016},
}
Figure