Evaluation of conditional treatment effects of adjuvant treatments on patients with synovial sarcoma using Bayesian subgroup analysis

Sung Wook Seo, Jisoo Kim, Jihye Son, Sungbin LimBMC Medical Informatics Decis. Mak 2020 · DOI · OpenReview

Introduction

Synovial sarcoma (SS) is a rare and aggressive soft-tissue cancer, accounting for approximately 6% of all such cases. Due to its rarity, conducting large-scale randomized controlled trials (RCTs)—the gold standard for medical evidence—is nearly impossible. Consequently, the efficacy of adjuvant chemotherapy and radiation therapy (RT) remains a subject of intense debate among oncologists. This study utilizes advanced statistical modeling to identify specific patient subgroups that truly benefit from these treatments, moving beyond deceptive population averages to provide a framework for precision oncology.

Methodology: Addressing Selection Bias and Heterogeneity

The primary challenge in analyzing retrospective observational data is "selection bias," where clinicians typically administer aggressive treatments to the most severely ill patients. To mitigate this and uncover hidden treatment patterns, the researchers combined two robust statistical frameworks.

Propensity Score Matching (PSM)

PSM was employed to simulate a randomized environment by pairing treated patients with untreated counterparts who shared nearly identical baseline characteristics, such as age, tumor size, and stage. This process balances the distribution of observed covariates between groups, allowing for a more accurate estimation of treatment effects by comparing "like with like."

Figure 4: (a) Standardized mean differences of covariates before (red dots) and after (black crosses) matching. (b) Kaplan-Meier survival curves for the entire cohort. (c) Kaplan-Meier survival curves for the matched cohort, showing the reversal of the apparent treatment effect after accounting for selection bias.

As illustrated in Figure 4a, the standardized mean differences for several key variables were significantly reduced after matching (moving from red dots to black crosses), indicating a well-balanced cohort. The impact of this correction is visible in the survival curves: while the raw data (Figure 4b) suggests that treated patients (red line) have lower survival due to being sicker initially, the matched data (Figure 4c) reveals that chemotherapy actually provides a survival advantage once the initial health disparity is accounted for.

Bayesian Subgroup Analysis

The core innovation of this study is the use of Bayesian subgroup analysis. The population was divided into 512 potential subgroups based on combinations of nine clinical variables. The researchers calculated the Conditional Treatment Effect (CTE) for each subgroup using a Beta-Binomial model, which updates prior beliefs with observed data to determine the "Win Probability"—the likelihood that a treated patient outlives an untreated patient within the same subgroup. Bayes Factors were then used to identify "credible" subgroups where the evidence of benefit was statistically substantial.

Key Insights: The Deception of Averages

The study's findings highlight that average treatment effects often mask critical variations in patient response. While chemotherapy showed an overall survival benefit (p=0.0001), this was largely driven by specific high-risk groups.

  • Chemotherapy's Optimal Subgroup: The most significant benefit was observed in patients older than 20, male, with large tumors (>5cm) in the extremities, advanced stage disease (SEER 3), and spindle cell histology. In this subgroup, untreated patients faced a 5.8-fold higher risk of death compared to those who received chemotherapy.
  • Radiation Therapy Nuances: Radiation did not show a significant overall benefit, yet it was highly effective for a specific early-stage subgroup: patients older than 20, male, with large tumors in the extremities, but with biphasic histology. Here, untreated patients had a 5.1-fold higher risk of death.
  • Potential for Harm: Strikingly, the model identified subgroups where chemotherapy was associated with a 9-fold increase in the risk of death, particularly in older males with early-stage tumors in the trunk or pelvis. This suggests that for certain low-risk profiles, treatment toxicity may outweigh marginal clinical benefits.

Validation and Limitations

The findings were validated using an external cohort of 242 Korean patients from three major institutions, which confirmed the survival trends observed in the primary SEER database (1,537 patients). This cross-population validation lends significant weight to the model's generalizability across different ethnicities.

However, the study is limited by the granularity of the SEER database, which lacks details on specific chemotherapy regimens and genetic mutations. Additionally, some of the 512 subgroups contained few patients, leading to wider credible intervals and increased uncertainty for those specific categories.

Conclusion and Future Directions

This research demonstrates that adjuvant treatment for synovial sarcoma should be highly personalized. By identifying specific high-risk and low-risk subgroups, clinicians can better navigate the trade-offs between treatment efficacy and toxicity. The authors suggest these Bayesian methods could be used to design "adaptive clinical trials," where the study focus shifts in real-time toward the subgroups showing the most benefit, potentially reducing the number of participants needed to reach definitive conclusions for rare diseases.

Summary updated

Abstract

The impact of adjuvant chemotherapy or radiation therapy on the survival of patients with synovial sarcoma (SS), which is a rare soft-tissue sarcoma, remains controversial. Bayesian statistical approaches and propensity score matching can be employed to infer treatment effects using observational data. Thus, this study aimed to identify the individual treatment effects of adjuvant therapies on the overall survival of SS patients and recognize subgroups of patients who can benefit from specific treatments using Bayesian subgroup analyses. We analyzed data from patients with SS obtained from the surveillance, epidemiology, and end results (SEER) public database. These data were collected between 1984 and 2014. The treatment effects of chemotherapy and radiation therapy on overall survival were evaluated using propensity score matching. Subgroups that could benefit from radiation therapy or chemotherapy were identified using Bayesian subgroup analyses. Based on a stratified Kaplan–Meier curve, chemotherapy exhibited a positive average causal effect on survival in patients with SS, whereas radiation therapy did not. The optimal subgroup for chemotherapy includes the following covariates: older than 20 years, male, large tumor (longest diameter > 5 cm), advanced stage (SEER 3), extremity location, and spindle cell type. The optimal subgroup for radiation therapy includes the following covariates: older than 20 years, male, large tumor (longest diameter > 5 cm), early stage (SEER 1), extremity location, and biphasic type. In this study, we identified high-risk patients whose variables include age (age > 20 years), gender, tumor size, tumor location, and poor prognosis without adjuvant treatment. Radiation therapy should be considered in the early stages for high-risk patients with biphasic types. Conversely, chemotherapy should be considered for late-stage high-risk SS patients with spindle cell types.
BibTeX
@article{seo2020evaluation,
  author = {Sung Wook Seo and Jisoo Kim and Jihye Son and Sungbin Lim},
  title = {{Evaluation of conditional treatment effects of adjuvant treatments on patients with synovial sarcoma using Bayesian subgroup analysis}},
  journal = {BMC Medical Informatics Decis. Mak.},
  year = {2020},
  doi = {10.1186/s12911-020-01305-9},
}
Figure