Trends in Female Authorship in Major Journals of 3 Oncology Disciplines, 2002-2018

Key Points Question How has the representation of women as authors in the oncology literature changed over time? Findings This cross-sectional study of 13 general oncology/medicine, radiation oncology, and surgical journals found that female authorship of oncology research articles rose from 25.5% in 2002 to 31.7% in 2018, with considerable variation based on author position, article type, and journal type. This rise is less than the increase in female oncology faculty over that time period. Meaning While female authorship in oncology research literature has increased over time, it has not kept up with the increase in female oncology faculty.


Introduction
Despite gender parity in medical school participation in recent years, 1,2 gender disparity remains in the numbers of practicing physicians, with variations by country 3 and specialty. 4Within academic centers in the United States, these discrepancies become more apparent with increasing academic rank, 5 reflecting "the leaky pipeline" of academic medicine.
Successful academic careers often hinge on the publication of peer-reviewed research.
0][11] Additionally, women are underpaid relative to their male counterparts within academia and receive smaller start-up packages, [12][13][14] receive less mentorship 15 and sponsorship, 16 face overt discrimination and harassment, 17 and often make personal or professional sacrifices because of traditional gender roles. 18,19though there has been an upward publishing trend in medicine, the proportion of female authors in the medical literature varies greatly by field, authorship position, and journal. 20Fields such as obstetrics and gynecology and pediatrics generally have higher rates of female authorship. 211][22] Different journals have varying rates of female first-authorship, 23 and women are underrepresented as presenters at major medical conferences. 24 general, female authorship in oncology has been increasing, 21 again with differences based on author position and discipline. 25The International Journal of Radiation Oncology, Biology, Physics (IJROBP), the radiation oncology journal with the highest impact factor and the most studied oncology journal, showed an increase in female authorship from 1980 through 2012 that corresponds with the rise in female full-time radiation oncology faculty and residents. 26ough previous work has examined trends in female authorship by sampling articles from a few years in a small number of journals, this study examines a more comprehensive data set: the entirety of the Medline-indexed oncology literature in high impact factor medical journals over a 17year span.

Data Collection
Data for this cross-sectional bibliometric analysis were collected from Medline via Medical Subject Headings (MeSH) term searches built in legacy PubMed.Thirteen clinically oriented journals with high impact factors for their respective fields and whose Medline citations included full first names were individually searched for oncology-specific articles.These journals were the Journal of the American Medical Association (JAMA), the Lancet, the New England Journal of Medicine, JAMA Oncology, Lancet Oncology, Journal of Clinical Oncology, IJROBP, Radiotherapy and Oncology, Practical Radiation Oncology, Annals of Surgical Oncology, Journal of Surgical Oncology, and Annals of Surgery.For each journal, separate searches were performed for each of 5 article types: clinical trials, observational studies (excluding case reports), systematic reviews with or without meta-analysis, general reviews, and other articles including letters, correspondences, news, replies, comments, and editorials (see eAppendix in the Supplement for exact search terms used).Analysis was restricted from 2002 through 2018 because author first names were not regularly included in Medline prior to 2002 and many of the included journals did not have all of their articles indexed with assigned MeSH terms for 2019 and later.Full lists of Medline citations for all resulting articles were directly exported from PubMed.No other exclusion criteria were applied.

Data Categorization
Authors for each extracted article were categorized as a first author, second author, last author, or other author based on their position in the extracted citation.Articles with a sole author had the author designated as a first author only, while those with 2 authors had the authors designated as a first or last author only.Last authorship was used as a surrogate for senior authorship because we were unable to efficiently extract the name of the corresponding author from the numerous articles included in the study.Journals were grouped either into general oncology, radiation oncology, or surgical oncology.Articles were categorized into 3 main groups.Primary articles included clinical trials and observational studies, such as cohort or cross-sectional studies.Secondary articles were reviews, systematic reviews, and meta-analyses.The remaining articles-including letters, correspondences, and replies-were grouped together as other articles.First names were used to infer authors' genders based on societal naming norms via the third-party service Gender-API.com(Markus Perl).Using their database comprising names from 191 different countries and algorithm, names were assigned as either male or female with an estimated accuracy score ranging from 50 through 100, with 50 signifying complete uncertainty and 100 signifying complete certainty in a name's gender.This service was used because it outperformed similar name-to-gender platforms in culturally diverse data sets. 27

Data Validation and Statistical Analysis
Data were analyzed between April and May 2020.All data analysis and statistical testing was performed in R version 3.6.1 (R Foundation for Statistical Computing) with a 2-tailed significance level prespecified at α = .05.Adjustments for multiple comparisons were not performed.Though the Gender-API.comservice has been previously validated, 27 internal validation was performed.Authors were randomly chosen until 500 were successfully assigned genders manually based on Google searches masked to the algorithmically assigned gender.Accuracy scores were scaled such that 0 and 100 corresponded to absolute certainty of male or female gender, respectively, with the midpoint representing complete uncertainty.Simple logistic regression was performed using the scaled accuracy scores to predict the manually assigned author genders.A receiver operating characteristic curve was generated from the model, and Youden's index was used to find the optimal cutoff of the accuracy score.
For the entire data set, percentages of female authorship by journal specialty, publication type, and authorship position were tabulated for each year.Comparisons between groups were performed using a Pearson χ 2 test of independence.The subset of primary journal articles (clinical trials and observational studies) were divided by discipline (general, radiation oncology, or surgical oncology), and multivariable logistic regression was performed with publication year, authorship position, and primary article type, as well as all 2-way and 3-way interactions, as independent variables to predict female authorship.Publication year was entered as years after 2002, and all categorical variables were dummy-coded.The Box-Tidwell test was used to confirm linearity between year and the log-odds of female authorship.

Data Summary and Validation
A total of 420 526 authors from 58 368 articles were found over the 17 years spanning 2002 through 2018.Of those authors, 400 945 were assigned a gender based on their name through Gender-API.com.Among the 19 581 authors for whom a gender was not assigned, 15 469 (79.0%) had only a first initial present in their Medline reference.Overall, Gender-API performed well, with a concordance index of 0.994 (95% CI, 0.989-0.998)between its scaled accuracy score and manually assigned author genders.The optimal cutoff was found to be at a scaled accuracy of 49, corresponding to correctly identifying 349 of 359 men (97.2%) and 136 of 141 women (96.5%).

Primary Literature
With the overall rise in female authorship for oncology-related journal articles, there were notable differences based on the specialty, publication type, and authorship position.The initial focus of this analysis was on primary research articles because of their relatively higher importance for career advancement.Figure 1 shows the percentage of female authors by discipline and authorship position vs year for clinical trials and observational studies.Each subgroup showed increased female authorship with each increasing year and the full results of the corresponding logistic regression models can be found in Table 3.
Among all articles published in general oncology journals, second authors were more likely to be women than first authors (Odds ratio [OR], 1.18; 95% CI, 1.03-1.35),whereas last authors were less likely to be women than first authors (OR, 0.60; 95% CI, 0.52-0.69).Women were less likely to be authors of clinical trials at each authorship position than authors at that respective position for observational studies (first, second, and last authors: OR, 0.58; 95% CI, 0.49-0.67;other authors: OR, 0.84; 95% CI, 0.79-0.89).For observational studies, other authors were less likely to women as a Three-year spans were used to help mitigate any year-to-year variation in comparing the beginning and end of this time frame, although presentation in this way blunts the small increase in female authorship within each of the 2 3-year spans.
b Other published works includes letters, correspondences, news, replies, comments, and editorials.

Secondary Literature and Other Articles
Figure 2 shows the overall percentage of female authors for secondary literature and other published articles for each discipline.The percentage of authors who were female at the beginning and end of the studied time span for these article types stratified by discipline can be seen in Table 2.The results of our study for those 2 years (Figures 1 and 2) are similar to these previous works, with small differences likely due to the averaging of multiple journals in this analysis.
Across all disciplines, last authors of primary articles were less likely to be women than first authors, suggesting fewer women in senior positions.This discrepancy was seen in other studies 20,21,25,26 and is not surprising given the surrounding academic framework.Chowdhary et al 28 demonstrated that among all accredited training programs in the US in early 2020, 37.1% of medical oncology faculty, 30.7% of radiation oncology faculty, and 38.8% of surgical oncology faculty were women.There were even fewer women in chair positions, at 21.7%, 11.7%, and 3.8%, respectively.

Figure 1 .F
Figure 1.Female Authorship Over Time in Primary Oncology Articles by Subspecialty

Figure 2 .B
Figure 2. Female Authorship Over Time in Secondary and All Other Oncology Articles by Subspecialty

Table 1 .
Trends in Female Authorship in Major Journals of3 Oncology Disciplines, 2002-2018The breakdown of the number of authors for whom a gender was assigned and the number of corresponding articles by discipline and publication type is shown in In total, 29.5% (95% CI, 29.4%-29.6%) of authors were identified as female, rising from 25.5% (95% CI, 24.7%-26.3%) in 2002 to 31.7% (95% CI, 31.2%-32.3%) in 2018.The most common article type in the general oncology journals were clinical trials, closely followed by observational studies.In contrast, the most common article types in both radiation oncology and surgical oncology journals were observational studies.The percentage of female authors over 3-year spans at the beginning and end of the studied time frame, stratified by publication type and discipline, can be seen in JAMA Network Open.2021;4(4):e212252.doi:10.1001/jamanetworkopen.2021.2252(Reprinted) April 6, 2021 3/10 Downloaded From: https://jamanetwork.com/ on 09/16/2023

Table 2 .
General oncology

Table 1 .
Total Number of Articles and Authors in Each Journal Type, Categorized by Publication Type

Table 2 .
Percentage of Female Authors Over 3-Year Spans at Beginning and End of Study Period, Stratified by Discipline and Study Type a Trends in Female Authorship in Major Journals of 3 Oncology Disciplines, 2002-2018 OR, 0.89; 95% CI, 0.80-0.98),thoughfor trials, other authors were more likely to be women than first authors (OR, 1.29; 95% CI, 1.13-1.47).For articles published in radiation oncology journals, last authors were less likely to be women than first authors, though the relative gap appeared to narrow with time (2002: OR, 0.36; 95% CI, 0.31-0.42vs 2018: OR, 0.57; 95% CI, 0.53-0.60).For observational studies, there was no difference in the odds of female authorship between second and first authors, though other authors were less JAMA Network Open.2021;4(4):e212252.doi:10.1001/jamanetworkopen.2021.2252(Reprinted) April 6, 2021 4/10 Downloaded From: https://jamanetwork.com/ on 09/16/2023 compared with first authors ( Trends in Female Authorship in Major Journals of3 Oncology Disciplines, 2002-2018difference in the odds of other authors being women as compared with first authors (OR, 1.11; 95% CI, 0.90-1.37).In surgical oncology journals, last authors were again less likely to be women than first authors (OR, 0.61; 95% CI, 0.50-0.73).Second authors were more likely to be women than first authors early in the timeframe, although the discrepancy disappeared over time (2002: OR 1.31, 95% CI, 1.11-1.55 in 2002 vs 2018: OR, 0.99; 95% CI, 0.94-1.05).Similarly, other authors were more likely to be women JAMA Network Open.2021;4(4):e212252.doi:10.1001/jamanetworkopen.2021.2252(Reprinted) April 6, 2021 5/10 Downloaded From: https://jamanetwork.com/ on 09/16/2023 26d faculty members over this time span.These estimates come from Association of American Medical Colleges data,4as well as 2 analyses by Chowdhary et al28and Ahmed et al.29The percentage of female hematology oncology trainees has increased from 38.5% in 2002 to 42.9% in 2017, although most years past 2010 have seen more than 45% female trainees.The percentage of female hematology-oncology faculty has also seen an increase from 28.7% in 2002 to 37.1% in 2018, with a peak of 38.8% in 2015.Overall, despite the approximately 9% to 10% increase in female hematology oncology faculty and the relative persistence of more than 45% female hematology-Dalal et al25described similar upward trends in the oncology literature.They analyzed all original articles, letters, notes, editorials, reviews, and proceedings in 5 major oncology journals at 4 different time points between 1990 and 2017.Among first authors in 2017, they found approximately 31% in the Annals of Surgical Oncology were women, approximately 34% in the IJROBP and Journal of Clinical Oncology, approximately 37% in JAMA Oncology, and approximately 45% in Cancer.For each journal that year, the percentage of female senior authors was nearly 10% lower than that for first authors.Ahmed et al26examined trends in female authorship in the IJROBP through 2012.For all original articles in 2012, they found 29.7% of first authors and 22.6% of senior authors were women. trainees