In RCTs, SES, defined as the primary endpoint, has some advantages over OS, as it requires a small sample size, a shorter follow-up period for the assessment, and low cost.12.21. Additionally, EFS is generally unaffected by post-progression treatments. Nevertheless, there is a downside to being affected by the method used to assess and determine biases, such as knowledge of therapy received22. It is possible that new diagnostic methods or anti-tumor agents would dilute the surrogacy relationship in the future, even if the current surrogacy would be strong23. Some reasons for the expectation of EFS as a surrogate endpoint for OS are social demand for early approval of new drugs and the theory that EFS is not just a surrogate endpoint for OS. ‘OS, but also makes positive contributions to patients’ quality of life21. Although many surrogate endpoint analyzes have evaluated EFS surrogacy for SG, a true endpoint in RCTs for various malignancies, there have been no such studies for RMS. This is the first report on surrogate endpoint analyzes for RMS and surrogacy analysis of intermediate endpoints for OS.
The control arms of the studies received several cytotoxic agents, including vincristine and actinomycin, while the experimental arms received more toxic and multidisciplinary treatments, additional radiation therapy or anti-tumor agents. In the meta-analysis, the experimental arms were more favorable than the control arms in terms of SES at 1 year. However, longer endpoints such as EFS at 3 and 5 years and OS at 3 and 5 years did not show superiority of experimental arms in RCTs conducted from 1974 were collected. In other words, for very early-stage EFS, the experimental arms may be more favorable than the control arms, but for later-stage EFS, the treatment effect between the arms would decrease. None of the new drug treatments have been successful. Overall, for RMS, there was only one RCT included in Table 1 showing that the OS of the experimental arm was superior to that of the control arm. It is difficult to carry out clinical trials due to a lack of effective drugs.
There are distinct prognoses and clinical courses between patients with localized RMS and those with metastatic RMS, even though these two groups of patients typically receive VAC-based chemotherapy. Therefore, we performed a sensitivity analysis excluding patients with metastatic RMS. After excluding metastatic patients with RMS, EFS HR was more weakly correlated with OS HR in the sensitivity analysis than in the primary analysis (Table 5). Similarly, the correlation between EFS at 1, 3 and 5 years and OS HR became progressively weak. In contrast, the correlation between OS at 1, 3, and 5 years and SG OS improved accordingly. This result is similar to that previously reported for osteosarcoma12 and Ewing’s sarcoma24.
In general, it is considered that the longer the survival time after recurrence, the lower the substitution of DFS for OS.25. Broglio et al.11 reported that if there was a small difference in SPP between the two treatment arms, longer periods of SPP would weaken the association between SSP HR and OS HR and make it difficult to demonstrate a statistically significant difference in OS between the treatment arms. Zer et al.26 assessed published RCTs of locally advanced/metastatic STS, with systemic treatment in at least one arm. They ultimately identified 52 RCTs eligible for analysis, including phase II and III studies, as in this study26. There were some differences from our study, i.e. they included different lines of treatment and various soft tissue sarcomas, and allowed for control arms such as placebo or best supportive care.26. Statistical analysis of the association between two parameters was assessed using a linear regression weighted by the size of the study sample and the strength of the association was assessed using the coefficient β standardized, not using the coefficient of determination R226. Their result showed significant correlations between PFS HR and OS HR and between RR OR and OS HR. The authors considered shorter SPP, around 12 months, to account for a large proportion of eligible studies, which likely led to significant correlations between HR SSP and HR OS26. On the other hand, as recommended by Schürmann et al.20, we used random-effects meta-regression. The study primarily included patients with ERMS, who typically have a longer PPS than those with other soft tissue sarcomas, as the 5-year survival rate of patients with ERMS after relapse is 20 at 52%.27. For this reason, the substitution of EFS for OS HR might have become weak in this study.
The relapse rate in patients with RMS who achieved complete remission or a stable mass was 31.1-36.2%28.29. In addition, recurrence occurred within 18 months of first diagnosis in 50.4% to 67.5% of relapsed patients.28.29 and within 5 years of first diagnosis in 95% of relapsed patients28. The 5-year survival rates after relapse in patients with EMS groups I and II/III were 52% and 20%, respectively.27. However, the 5-year survival rates after relapse in patients with ARMS groups I and II-IV or undifferentiated sarcoma were 40% and 3%, respectively.27. In addition, the time to relapse after the end of primary treatment has a significant influence on the prognosis. Four-year survival rates after relapse within 6, 6-12, and more than 12 months were 12%, 21%, and 41%, respectively30. According to these results, ERMS with post-relapse survival was dominant because 11 RCTs used in this study included 53.9% ERMS and 29.6% ARMS; therefore, the substitution of EFS for OS HR was found to be weak. EFS at 5 years was weaker correlated with OS HR than EFS at 3 years, possibly because patients with early relapse RMS tend to die early and patients with relapse late tend to have long survival after relapse. Another reason is that individual post-relapse treatments were heterogeneous. Some relapsed patients may be able to undergo complete surgical excision and have long-term post-relapse survival29.
DFS has not yet been validated as a surrogate measure of OS in patients with breast cancer. Nevertheless, when patients with HER2-positive early breast cancer are selected as the object, DFS might be an acceptable surrogate for OS. 31. Goldberg et al. 22 suggested that an appropriate biomarker, clinically defined patient selection, and receipt of effective treatment would likely lead to restoration of PFS surrogacy for SG. In the case of RMS, which is known to be an ultra-rare sarcoma32, it seemed difficult to apply patient selection and specific treatment for RMS RCTs. Thus, in terms of malignancies, it is important to discover or develop new biomarkers that can be acceptable surrogates for SG.
Our study has several limitations. First, we did not use individual data but published data. Second, most eligible RCTs did not describe the study phase. Third, only two RCTs described ITT analysis. Finally, there were differences between each follow-up examination or subsequent RCT treatment because eligible patients were registered for a long period (1979-2016).
This study concludes that when SES is considered the primary endpoint of an RCT of RMS, a follow-up period of at least three years is required. Moreover, this result shows that the association between EFS and OS was modest, and surrogacy EFS for OS in RCTs of RMS was not confirmed. Therefore, there is a need to discover or develop new biomarkers for RMS that can be an acceptable surrogate for OS.