A social network analysis of treatment discoveries in cancer

• Published in: PloS one Vol. 6; no. 3; p. e18060
• Main Authors: Tsalatsanis, Athanasios, Barnes, Laura, Hozo, Iztok, Skvoretz, John, Djulbegovic, Benjamin
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 28-03-2011; Public Library of Science (PLoS)
• Subjects: Analysis; Cancer; Cancer treatment; Clinical trials; Clustering; Collaboration; Drug therapy; Ethics; Evidence-based medicine; Health aspects; Historicism; Humans; Information sources; Interaction models; Law; Library associations; Life expectancy; Life span; Medical libraries; Medical research; Medicine; Meta-analysis; Neoplasms - therapy; Network analysis; Network centrality; Normal distribution; Oncology; Peer relationships; Power law; Primary care; Randomized Controlled Trials as Topic; Researchers; Social and Behavioral Sciences; Social factors; Social interaction; Social interactions; Social network analysis; Social networks; Social organization; Social relations; Social Support; Sociology; Solid tumors; Studies; Success; Survival; Survival Analysis; Treatment methods; Treatment Outcome; Tumors; Indiana; Florida; United States > US
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0018060

Controlled clinical trials are widely considered to be the vehicle to treatment discovery in cancer that leads to significant improvements in health outcomes including an increase in life expectancy. We have previously shown that the pattern of therapeutic discovery in randomized controlled trials (RCTs) can be described by a power law distribution. However, the mechanism generating this pattern is unknown. Here, we propose an explanation in terms of the social relations between researchers in RCTs. We use social network analysis to study the impact of interactions between RCTs on treatment success. Our dataset consists of 280 phase III RCTs conducted by the NCI from 1955 to 2006. The RCT networks are formed through trial interactions formed i) at random, ii) based on common characteristics, or iii) based on treatment success. We analyze treatment success in terms of survival hazard ratio as a function of the network structures. Our results show that the discovery process displays power law if there are preferential interactions between trials that may stem from researchers' tendency to interact selectively with established and successful peers. Furthermore, the RCT networks are "small worlds": trials are connected through a small number of ties, yet there is much clustering among subsets of trials. We also find that treatment success (improved survival) is proportional to the network centrality measures of closeness and betweenness. Negative correlation exists between survival and the extent to which trials operate within a limited scope of information. Finally, the trials testing curative treatments in solid tumors showed the highest centrality and the most influential group was the ECOG. We conclude that the chances of discovering life-saving treatments are directly related to the richness of social interactions between researchers inherent in a preferential interaction model.