Collaboration between researchers has been shown to influence their productivity and scientific impact. Although these ties have been widely discussed in the literature, the nature of the co-authorship networks between the most successful scholars remains a question. To provide an answer, this study conducts a cross-case analysis of the collaboration networks between Highly Cited Researchers, focusing on the research output and co-authorship patterns in Highly Cited Papers across three award categories: Clinical Medicine, Materials Science, and Social Sciences. Our findings indicate that there are category-specific differences in publication output and the intensity of collaboration between Highly Cited Researchers. Notably, Highly Cited Researchers in the Social Sciences demonstrate a less collaborative approach to research than those in Clinical Medicine and Materials Science. While Highly Cited Researchers in all three categories featured interconnected collaboration networks among themselves, those in Clinical Medicine and Materials Science exhibited a more collaborative environment, while those in Social Sciences showed a tendency towards independent research efforts. The case of Social Sciences is further evidenced by higher fragmentation within the collaboration network of Social Sciences, indicating a less cohesive collaborative framework. The analysis of the Giant Component—the largest cohesive subset of the network—revealed that it is less representative of the overall network structure in the Social Sciences than in Clinical Medicine and Materials Science. Finally, the centrality measures indicated that Highly Cited Researchers with high betweenness and closeness centrality act as crucial bridges within each network, significantly shaping the structural cohesion and collaborative dynamics of their respective fields.

Background: To our knowledge, no previous systematic review and meta-analysis of doping prevalence in sport from indirect estimation models (IEM) exists. Objective: To conduct a systematic review and meta-analysis of empirical IEM-based studies of admitted doping prevalence in sport. Methods: We conducted electronic database and ad hoc searches up to March 2025, and estimated lifetime and past year prevalence rates through a cross-classified model including prevalence (lifetime vs. past year), sample (competitive vs. recreational) and sports (multi-sport vs. single-sport) types. Results: Forty-six records (K) were included in the review (k [subset records included in the meta-analysis] = 30, n [independent studies from the records] = 34). The World Anti-Doping Agency’s definition of doping use was applied for data collection in most studies (k = 18), and doping prevalence was mostly assessed as past year/season (k = 20). Studies included in the meta-analysis were mostly conducted in Europe (k = 22) and applied the Unrelated Question (k = 8) and Forced Response with Cheater Detection (k = 6) models. Study participants were mostly multi-sport (k = 20) and competed at diverse levels, and most data (k = 28) was collected outside sport events. The corpus included articles that re-analysed existing data (k = 4). Lifetime prevalence was highest for multi-sport competitive athletes (22.6%) and lowest for single-sport competitive athletes (12.7%), whereas past year prevalence was highest for single-sport recreational sportspersons (15.5%) and lowest for multi-sport recreational sportspersons (8.7%). Conclusions: Under IEM, about one of five multi-sport competitive athletes admitted to ever doping whereas about one of six of single-sport recreational sportspersons admitted to doping in the past year. Furthermore, multi-sport (vs. single-sport) competitive athletes show relatively higher doping prevalences, whereas single-sport (vs. multi-sport) recreational sportspersons report relatively higher doping prevalences. Secondary (re-)analysis presents a novel methodological challenge for meta-analyses. Registration PROSPERO: CRD42022373691.

Interpreting doping prevalence estimates generated through indirect estimation models (IEMs) remains challenging for sport policy and governance due to the wide variation in reported rates and methodological complexity. In this study, we combined a critical appraisal of the methodological and epistemic developments of IEM applications to doping prevalence with a bibliometric analysis of publication trends, citation patterns, and collaboration networks, using a convergent parallel mixed-methods design. Across 52 records published between 2002 and 2026, this study maps the scientific landscape of IEM-based doping prevalence research. Findings show that IEM-based prevalence research is methodologically sophisticated yet institutionally dispersed and largely Eurocentric, reflecting a field still consolidating its standards and disciplinary identity. Over time, the focus has shifted from reporting prevalence rates to methodological critique and re-analysis of existing datasets. Reported prevalence estimates, ranging from 0 to 57.1%, are highly sensitive to modelling assumptions about athlete behaviour in complex survey environments. While this trend strengthens rigour, it also complicates evidence synthesis for policy actors and risks undermining trust in IEM-based estimates if poorly communicated. Anti-doping organisations and researchers should treat IEM-derived prevalence as bounded indicators rather than definitive rates and integrate prevalence evidence with contextual data for transparent policy and public communication.