Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.
- MeSH
- big data * MeSH
- glioblastom * MeSH
- lidé MeSH
- šíření informací MeSH
- strojové učení MeSH
- vzácné nemoci MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
- Research Support, N.I.H., Extramural MeSH
Over the past decade, wingless-activated (WNT) medulloblastoma has been identified as a candidate for therapy de-escalation based on excellent survival; however, a paucity of relapses has precluded additional analyses of markers of relapse. To address this gap in knowledge, an international cohort of 93 molecularly confirmed WNT MB was assembled, where 5-year progression-free survival is 0.84 (95%, 0.763-0.925) with 15 relapsed individuals identified. Maintenance chemotherapy is identified as a strong predictor of relapse, with individuals receiving high doses of cyclophosphamide or ifosphamide having only one very late molecularly confirmed relapse (p = 0.032). The anatomical location of recurrence is metastatic in 12 of 15 relapses, with 8 of 12 metastatic relapses in the lateral ventricles. Maintenance chemotherapy, specifically cumulative cyclophosphamide doses, is a significant predictor of relapse across WNT MB. Future efforts to de-escalate therapy need to carefully consider not only the radiation dose but also the chemotherapy regimen and the propensity for metastatic relapses.
- Publikační typ
- časopisecké články MeSH
While molecular subgrouping has revolutionized medulloblastoma classification, the extent of heterogeneity within subgroups is unknown. Similarity network fusion (SNF) applied to genome-wide DNA methylation and gene expression data across 763 primary samples identifies very homogeneous clusters of patients, supporting the presence of medulloblastoma subtypes. After integration of somatic copy-number alterations, and clinical features specific to each cluster, we identify 12 different subtypes of medulloblastoma. Integrative analysis using SNF further delineates group 3 from group 4 medulloblastoma, which is not as readily apparent through analyses of individual data types. Two clear subtypes of infants with Sonic Hedgehog medulloblastoma with disparate outcomes and biology are identified. Medulloblastoma subtypes identified through integrative clustering have important implications for stratification of future clinical trials.
- MeSH
- genomika MeSH
- individualizovaná medicína * MeSH
- kohortové studie MeSH
- lidé MeSH
- meduloblastom klasifikace genetika terapie MeSH
- metylace DNA MeSH
- shluková analýza MeSH
- stanovení celkové genové exprese MeSH
- variabilita počtu kopií segmentů DNA MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
BACKGROUND: Patients with incomplete surgical resection of medulloblastoma are controversially regarded as having a marker of high-risk disease, which leads to patients undergoing aggressive surgical resections, so-called second-look surgeries, and intensified chemoradiotherapy. All previous studies assessing the clinical importance of extent of resection have not accounted for molecular subgroup. We analysed the prognostic value of extent of resection in a subgroup-specific manner. METHODS: We retrospectively identified patients who had a histological diagnosis of medulloblastoma and complete data about extent of resection and survival from centres participating in the Medulloblastoma Advanced Genomics International Consortium. We collected from resections done between April, 1997, and February, 2013, at 35 international institutions. We established medulloblastoma subgroup affiliation by gene expression profiling on frozen or formalin-fixed paraffin-embedded tissues. We classified extent of resection on the basis of postoperative imaging as gross total resection (no residual tumour), near-total resection (<1·5 cm(2) tumour remaining), or sub-total resection (≥1·5 cm(2) tumour remaining). We did multivariable analyses of overall survival and progression-free survival using the variables molecular subgroup (WNT, SHH, group 4, and group 3), age (<3 vs ≥3 years old), metastatic status (metastases vs no metastases), geographical location of therapy (North America/Australia vs rest of the world), receipt of chemotherapy (yes vs no) and receipt of craniospinal irradiation (<30 Gy or >30 Gy vs no craniospinal irradiation). The primary analysis outcome was the effect of extent of resection by molecular subgroup and the effects of other clinical variables on overall and progression-free survival. FINDINGS: We included 787 patients with medulloblastoma (86 with WNT tumours, 242 with SHH tumours, 163 with group 3 tumours, and 296 with group 4 tumours) in our multivariable Cox models of progression-free and overall survival. We found that the prognostic benefit of increased extent of resection for patients with medulloblastoma is attenuated after molecular subgroup affiliation is taken into account. We identified a progression-free survival benefit for gross total resection over sub-total resection (hazard ratio [HR] 1·45, 95% CI 1·07-1·96, p=0·16) but no overall survival benefit (HR 1·23, 0·87-1·72, p=0·24). We saw no progression-free survival or overall survival benefit for gross total resection compared with near-total resection (HR 1·05, 0·71-1·53, p=0·8158 for progression-free survival and HR 1·14, 0·75-1·72, p=0·55 for overall survival). No significant survival benefit existed for greater extent of resection for patients with WNT, SHH, or group 3 tumours (HR 1·03, 0·67-1·58, p=0·89 for sub-total resection vs gross total resection). For patients with group 4 tumours, gross total resection conferred a benefit to progression-free survival compared with sub-total resection (HR 1·97, 1·22-3·17, p=0·0056), especially for those with metastatic disease (HR 2·22, 1·00-4·93, p=0·050). However, gross total resection had no effect on overall survival compared with sub-total resection in patients with group 4 tumours (HR 1·67, 0·93-2·99, p=0·084). INTERPRETATION: The prognostic benefit of increased extent of resection for patients with medulloblastoma is attenuated after molecular subgroup affiliation is taken into account. Although maximum safe surgical resection should remain the standard of care, surgical removal of small residual portions of medulloblastoma is not recommended when the likelihood of neurological morbidity is high because there is no definitive benefit to gross total resection compared with near-total resection. FUNDING: Canadian Cancer Society Research Institute, Terry Fox Research Institute, Canadian Institutes of Health Research, National Institutes of Health, Pediatric Brain Tumor Foundation, and the Garron Family Chair in Childhood Cancer Research.
- MeSH
- dítě MeSH
- dospělí MeSH
- kojenec MeSH
- kombinovaná terapie MeSH
- lidé MeSH
- magnetická rezonanční tomografie MeSH
- meduloblastom klasifikace genetika patologie chirurgie MeSH
- nádory mozku klasifikace genetika patologie chirurgie MeSH
- předškolní dítě MeSH
- přežití po terapii bez příznaků nemoci MeSH
- prognóza * MeSH
- progrese nemoci MeSH
- retrospektivní studie MeSH
- Check Tag
- dítě MeSH
- dospělí MeSH
- kojenec MeSH
- lidé MeSH
- mužské pohlaví MeSH
- předškolní dítě MeSH
- ženské pohlaví MeSH
- Publikační typ
- časopisecké články MeSH
- Geografické názvy
- Kanada MeSH
