Search

Abstract Background Machine learning (ML) has made a significant impact in medicine and cancer research; however, its impact in these areas has been undeniably slower and more limited than in other application domains. A major reason for this has been the lack of availability of patient data to the broader ML research community, in large part due to patient privacy protection concerns. High-quality, realistic, synthetic datasets can be leveraged to accelerate methodological developments in medicine. By and large, medical data is high dimensional and often categorical. These characteristics pose multiple modeling challenges. Methods In this paper, we evaluate three classes of synthetic data generation approaches; probabilistic models, classification-based imputation models, and generative adversarial neural networks. Metrics for evaluating the quality of the generated synthetic datasets are presented and discussed. Results While the results and discussions are broadly applicable to medical data, for demonstration purposes we generate synthetic datasets for cancer based on the publicly available cancer registry data from the Surveillance Epidemiology and End Results (SEER) program. Specifically, our cohort consists of breast, respiratory, and non-solid cancer cases diagnosed between 2010 and 2015, which includes over 360,000 individual cases. Conclusions We discuss the trade-offs of the different methods and metrics, providing guidance on considerations for the generation and usage of medical synthetic data.

Abstract Background While full disclosure of information on placebo control in participant information leaflets (PILs) in a clinical trial is ethically required during informed consent, there have been concerning voices such complete disclosures may increase unnecessary nocebo responses, breach double-blind designs, and/or affect direction of trial outcomes. Taking an example of acupuncture studies, we aimed to examine what participants are told about placebo controls in randomized, placebo-controlled trials, and how it may affect blinding and trial outcomes. Methods Authors of published randomized, placebo-controlled trials of acupuncture were identified from PubMed search and invited to provide PILs for their trials. The collected PILs were subjected to content analysis and categorized based on degree of information disclosure on placebo. Blinding index (BI) as a chance-corrected measurement of blinding was calculated and its association with different information disclosure was examined. The impact of different information disclosure from PILs on primary outcomes was estimated using a random effects model. Results In 65 collected PILs, approximately 57% of trials fully informed the participants of placebo control, i.e. full disclosure, while the rest gave deceitful or no information on placebo, i.e. no disclosure. Placebo groups in the studies with no disclosure tended to make more opposite guesses on the type of received intervention than those with disclosure, which may reflect wishful thinking (BI −0.21 vs. −0.16; p = 0.38). In outcome analysis, studies with no disclosure significantly favored acupuncture than those with full disclosure (standardized mean difference − 0.43 vs. −0.12; p = 0.03), probably due to enhanced expectations. Conclusions How participants are told about placebos can be another potential factor that may influence participant blinding and study outcomes by possibly modulating patient expectation. As we have few empirical findings on this issue, future studies are needed to determine whether the present findings are relevant to other medical disciplines and at the same time a routine practice of fully disclosing placebo information in PILs calls for reevaluation.

Background: Machine learning (ML) has made a significant impact in medicine and cancer research; however, its impact in these areas has been undeniably slower and more limited than in other application domains. A major reason for this has been the lack of availability of patient data to the broader ML research community, in large part due to patient privacy protection concerns. High-quality, realistic, synthetic datasets can be leveraged to accelerate methodological developments in medicine. By and large, medical data is high dimensional and often categorical. These characteristics pose multiple modeling challenges.Methods: In this paper, we evaluate three classes of synthetic data generation approaches; probabilistic models, classification-based imputation models, and generative adversarial neural networks. Metrics for evaluating the quality of the generated synthetic datasets are presented and discussed.Results: While the results and discussions are broadly applicable to medical data, for demonstration purposes we generate synthetic datasets for cancer based on the publicly available cancer registry data from the Surveillance Epidemiology and End Results (SEER) program. Specifically, our cohort consists of breast, respiratory, and non-solid cancer cases diagnosed between 2010 and 2015, which includes over 360,000 individual cases.Conclusions: We discuss the trade-offs of the different methods and metrics, providing guidance on considerations for the generation and usage of medical synthetic data. [ABSTRACT FROM AUTHOR]

Background: While full disclosure of information on placebo control in participant information leaflets (PILs) in a clinical trial is ethically required during informed consent, there have been concerning voices such complete disclosures may increase unnecessary nocebo responses, breach double-blind designs, and/or affect direction of trial outcomes. Taking an example of acupuncture studies, we aimed to examine what participants are told about placebo controls in randomized, placebo-controlled trials, and how it may affect blinding and trial outcomes.Methods: Authors of published randomized, placebo-controlled trials of acupuncture were identified from PubMed search and invited to provide PILs for their trials. The collected PILs were subjected to content analysis and categorized based on degree of information disclosure on placebo. Blinding index (BI) as a chance-corrected measurement of blinding was calculated and its association with different information disclosure was examined. The impact of different information disclosure from PILs on primary outcomes was estimated using a random effects model.Results: In 65 collected PILs, approximately 57% of trials fully informed the participants of placebo control, i.e. full disclosure, while the rest gave deceitful or no information on placebo, i.e. no disclosure. Placebo groups in the studies with no disclosure tended to make more opposite guesses on the type of received intervention than those with disclosure, which may reflect wishful thinking (BI -0.21 vs. -0.16; p = 0.38). In outcome analysis, studies with no disclosure significantly favored acupuncture than those with full disclosure (standardized mean difference - 0.43 vs. -0.12; p = 0.03), probably due to enhanced expectations.Conclusions: How participants are told about placebos can be another potential factor that may influence participant blinding and study outcomes by possibly modulating patient expectation. As we have few empirical findings on this issue, future studies are needed to determine whether the present findings are relevant to other medical disciplines and at the same time a routine practice of fully disclosing placebo information in PILs calls for reevaluation. [ABSTRACT FROM AUTHOR]